Most days, a fingertip oximeter is good enough for anyone. The day it actually matters — sick at home, deciding whether a borderline reading should send you to an ER — is the day the bias bites, and it bites darker-skinned readers harder. The knowledge costs nothing; the cost of not knowing has been counted in delayed COVID treatments and excess deaths.
Pulse oximetry works by passing red and infrared light through the finger and reading what comes out the other side. Oxygenated and deoxygenated blood absorb those two wavelengths in different ratios, and the device backs out the oxygen saturation from the math. The trouble is that melanin in the skin absorbs the same red wavelength the device is counting on — so on darker skin, less red light makes it through, and the algorithm shifts its estimate of oxygen upward even when the blood underneath is the same.
The average error is small — about one or two percentage points high Fawzy 2022. The reason it matters is that the important medical thresholds sit exactly there. 88% is the floor for "safe oxygen." 92% triggers supplemental oxygen on a hospital ward. 94% gated COVID-era steroids and antivirals. A one-or-two-point upward shift across those numbers is the difference between treatment and dismissal.
What the bias actually does to people
The lab finding was confirmed in real hospitals during COVID. Across a 178-hospital intensive-care dataset and the early-pandemic admissions at a large US academic centre, Black patients with "reassuring" pulse-oximeter readings in the 92–96% range were almost three times as likely to actually be dangerously low on oxygen as white patients with the same reading — 11.7% vs 3.6% rate of missed hypoxia Sjoding 2020. A larger multi-hospital analysis linking this hidden low-oxygen to outcomes found a 41% higher death rate among patients whose true hypoxia went undetected Wong et al. 2021. And the disparity isn't only in Black patients: Hispanic, Asian, and South Asian patients all show a measurable shift in the same direction, smaller but real Fawzy 2022.
How this actually plays out
The exact scene the literature documents: a dark-skinned family member is short of breath. The ER reads 93% on a finger clip. They're sent home with antibiotics. The next morning, when an arterial sample is finally drawn, the true oxygen turns out to have been in the low 80s — but by then they're in the ICU, or worse. Multiply that scene by the millions of pulse-oximeter readings taken every day during the pandemic, and you get the disparity the Hopkins team measured: hundreds of Black and Hispanic patients who never qualified, by the device's number, for drugs that would have saved them by their actual blood Fawzy 2022.
The home version is quieter. You're checking your own oxygen because you've got a cough and a friend mentioned that COVID can drop your saturation without warning. The reading says 95%. You stay in bed. You're a dark-skinned reader with a one-or-two-point average bias, so your actual oxygen is closer to 93 or 94 — still probably fine, but the comfortable safety margin you thought you had is half what it looked like. The bias doesn't kill you in that moment; it narrows the margin between you and a wrong call, every time the device is the thing telling you whether to act.
How to read the number, knowing this
The interpretation rule used by critical-care physicians who think about this every shift is short, and it works on whatever device you already own.
Two things to unlearn while you're at it. The first: pulse oximeters work the same on everyone. They don't, and they haven't for the 35 years the bias has been documented Jubran 1990. The second: the fix is a race-based correction in the software. Recent technical work explicitly rejects that — race is a proxy for skin colour, not the variable itself, and writing race into the algorithm just relocates the problem. The right fix is validating the hardware against actual skin tones, which the FDA is now finally requiring of new device submissions, plus the bedside heuristic above for every device already in service FDA 2025.
Where the bias bites in real life
Four predictable scenarios where the gap between "what the device says" and "what's actually happening in the blood" turns into a wrong call.
On a hospital ward. A patient with pneumonia or post-op respiratory weakness reads 92% on the clip, is judged stable, and isn't escalated to higher-flow oxygen. The true saturation is below 88%. The ICU study that quantified this found Asian, Black, and Hispanic patients received measurably less supplemental oxygen than white patients with the same true oxygen level Gottlieb 2022.
In the ER waiting room. A patient walks in short of breath; the triage clip reads 94%; they're queued or sent home. Their actual hypoxia is missed. In the Hopkins COVID cohort, Black patients were 29% less likely than white patients to be flagged for treatment by SpO2 Fawzy 2022.
At home during a respiratory illness. Someone with COVID or pneumonia checks their drug-store oximeter, sees 95%, decides to ride it out. Cheap consumer devices have larger errors stacked on top of the skin-tone bias, and the layperson doesn't have the symptom-and-trend instinct a clinician would weigh FDA 2021.
Before salvage therapy. Among the sickest patients on the brink of being placed on full lung-bypass machines (ECMO), the rate of undetected severe hypoxia in the 92–96% range was 21.5% in Black patients versus 10.2% in white — twice as many missed warnings at the moment when missing one is catastrophic Valbuena 2022.
Picking a device for home
Most fingertip oximeters sold in pharmacies and on Amazon — the $20–60 clips — are classified by the FDA as "wellness" products. That word is doing a lot of work: it means the agency does not review their accuracy at all FDA 2021. Independent testing of these has found wide variability and large absolute errors, even before the skin-tone bias compounds the problem. They are fine for casual curiosity — a hiker checking their oxygen at altitude, a curious morning check after a poor night's sleep. They are not the device you want when an illness might be turning into something serious.
For at-home use during a real concern — pneumonia, COVID, a flare of COPD or asthma in someone with a history — the right buy is one of the FDA-cleared over-the-counter medical pulse oximeters that came on the market in 2023–2024. Masimo's MightySat OTC and Nonin's TruO2 were both tested across the full range of skin tones to meet the FDA's tightened standard. They cost around $200–300 — substantially more than a drug-store clip, but in the same ballpark as a single ER copay.
For clinical-grade devices in hospitals, the FDA's 2025 draft guidance now demands larger and more skin-diverse validation cohorts before new oximeters are cleared FDA 2025. That doesn't retroactively fix the millions of devices already in service in hospitals today — which is why the interpretation rule above is the workable fix, not waiting for new hardware.
Related
- Sleep apnea testing. Home sleep tests rely on pulse oximetry to count overnight oxygen drops. The pigment bias propagates into how severe an apnea looks on paper, and is quietly under-studied.
- Wearable continuous oxygen tracking. Apple Watch, Oura, Garmin and Fitbit all advertise SpO2 as a wellness feature. They aren't medical devices and the bias on a wrist-worn reflectance sensor across skin tones is mostly uncharacterized.
- Carbon monoxide poisoning. A separate failure mode of pulse oximetry: the device reads completely normal even at lethal CO levels, on any skin tone. Worth knowing if you have a gas furnace.
- — If you monitor your oxygen at home for a lung condition, the skin-tone bias matters most exactly when you're sick and the number is borderline.
- — One blind spot to know: in carbon-monoxide poisoning the oximeter reads perfectly normal at lethal levels. A CO alarm is the real safeguard.
- — Another everyday medical tool that was quietly less accurate for some patients — same lesson about not trusting a single number blindly.
- — Same trap, different test: HbA1c also reads falsely off in some people. A clinical number isn't truth — know when yours is unreliable.
- — Same skill as reading any lab: the printed number isn't the truth. A borderline oximeter reading needs the same skeptical eye.
- — Home sleep tests count overnight oxygen dips with the same finger sensor — the pigment bias can quietly make your apnea look milder than it is.
Substance and claimed effects
The substance is pulse oximeter pigment bias: a systematic measurement error in which the peripheral oxygen saturation (SpO2) reported by a finger or earlobe pulse oximeter overestimates the true arterial oxygen saturation (SaO2) more on darker-pigmented skin than on lighter-pigmented skin. The clinical consequence is occult hypoxemia — a patient whose blood is truly hypoxic but whose oximeter reads in the green zone. This entry covers the mechanism, the controlled-lab and real-world evidence, the documented downstream consequences (delayed care escalation, lower supplemental oxygen administration, missed COVID-19 treatment eligibility, higher mortality), and the implications for at-home oximeter use during respiratory illness. Out-of-scope: pulse oximetry's other accuracy limitations (motion, nail polish, methemoglobinemia, carbon monoxide poisoning) and sleep-apnea oximetry as a primary use case, both signposted in §3b.out-of-scope.
Evidence per addressing question
Mechanism
Pulse oximetry estimates the ratio of oxygenated to total hemoglobin from the differential absorption of light at two wavelengths — typically red (660 nm) and infrared (940 nm) — transmitted through a fingertip or earlobe. Oxyhemoglobin and deoxyhemoglobin absorb the two wavelengths at different ratios, and the device backs out SpO2 from the pulsatile component of the transmitted signal. Melanin absorbs strongly in the red portion of the spectrum, overlapping with deoxyhemoglobin's absorption peak; in dark-pigmented skin, the oximeter underestimates the pulsatile-deoxyhemoglobin fraction and the algorithm shifts toward a higher SpO2 than the underlying blood justifies Bickler 2005.
The bias direction is consistent across studies: SpO2 over-reads SaO2 in dark-pigmented patients, with magnitude growing as true SaO2 falls — i.e., worst exactly when the reading matters most Bickler 2005. Compounding contributors include low peripheral perfusion (which co-varies with shock, hypothermia, vasopressors), pulsatile-signal degradation, and manufacturer calibration databases historically derived from majority-white volunteer cohorts FDA 2021. The 2022 FDA advisory panel concluded that skin pigmentation is a meaningful but not sole contributor to the disparate performance, with perfusion explaining additional variance FDA 2025.
Evidence
Controlled lab work. Bickler and colleagues at the UCSF Hypoxia Research Laboratory desaturated healthy volunteers to plateau SaO2 levels between 60% and 100% while comparing finger-probe SpO2 to co-oximetry. Three commercial oximeters all overestimated SaO2 in dark-pigmented subjects, with maximum bias of 3.56 ± 2.45 percentage points at SaO2 60–70%; bias was smaller but still present at higher saturations Bickler 2005. A 1990 bedside study had already shown that a 92% SpO2 target reliable in white ventilated patients was unreliable in Black patients, who needed 95% to guarantee adequate oxygenation Jubran and Tobin 1990 — the phenomenon has been in the literature for 35 years.
Real-world retrospective cohorts. Sjoding and colleagues analyzed paired SpO2/SaO2 measurements in a 178-hospital ICU dataset and at the University of Michigan during the early COVID-19 wave. Among patients with SpO2 92–96% (the "reassuring" range), arterial blood gas confirmed SaO2 <88% in 11.7% of Black-patient measurements vs 3.6% of white-patient measurements — an almost three-fold higher rate of missed hypoxia Sjoding 2020. Wong and colleagues replicated this across three EHR databases (eICU-CRD, MIMIC-III, MIMIC-IV), confirming higher hidden-hypoxemia rates in Asian, Black, and Hispanic patients vs white, and linking hidden hypoxemia to a 41% higher mortality risk plus measurable organ-dysfunction signal Wong et al. 2021.
Pre-ECMO cohort. Valbuena and colleagues studied patients about to undergo ECMO — i.e., on the edge of catastrophic respiratory failure. Pre-ECMO occult hypoxemia at SpO2 92–96% was 10.2% in white, 21.5% in Black (p=0.031 vs white), 8.6% in Hispanic, 9.2% in Asian patients Valbuena et al. 2022.
Stakes
The downstream care consequences are the strongest argument that this is a clinically actionable bias, not an academic curiosity. Fawzy and colleagues at Johns Hopkins examined >6,000 hospitalized COVID-19 patients and found pulse oximetry overestimated SaO2 by 1.2 percentage points in Black, 1.1 in Hispanic, 1.7 in Asian patients. Mean bias is small, but it sat across the SpO2 ≤94% threshold for dexamethasone and remdesivir eligibility: Black patients were 29% and Hispanic patients 23% less likely than white patients to be flagged for treatment by SpO2. Treatment was delayed by 5–7 hours in many; 451 patients (55% Black, 27% Hispanic) never qualified despite predicted-from-blood-gas eligibility Fawzy et al. 2022. Gottlieb's ICU cohort (n=3,069) showed Asian, Black, and Hispanic patients receiving significantly less supplemental oxygen for the same true SaO2 than white patients — the bias visible directly in the volume of O2 delivered Gottlieb et al. 2022.
Practice / clinical consensus
The Anesthesia Patient Safety Foundation and the American Thoracic Society have issued statements acknowledging the bias and urging clinicians to triangulate SpO2 with clinical signs, capnography, and ABG. The FDA issued a safety communication in 2021 warning that pulse oximeters may be less accurate in dark-pigmented skin, that wellness oximeters are not reviewed for accuracy, and that clinical decisions should not rest on SpO2 alone FDA 2021. The agency convened advisory panels in 2022 and 2024, and published draft guidance in January 2025 requiring validation cohorts of ≥150 subjects with ≥25% dark-pigmented participants, ≥30% of data points from darker-skinned individuals, objective skin-tone assessment via the Monk Skin Tone scale, and standardized reporting FDA 2025. Critically, the draft guidance applies only to oximeters marketed for medical purposes; "general wellness" OTC fingertip clips remain outside FDA accuracy review.
The clinical interpretation heuristic most widely used by critical-care physicians familiar with the bias: in a dark-pigmented patient with a borderline SpO2 (92–94%), mentally subtract 2–3 percentage points and lower the threshold for ABG, supplemental oxygen escalation, and ICU consultation. Symptoms and the trend over time should override any single reading.
Community / lay evidence
During the COVID-19 pandemic, online medical forums (Twitter clinician threads, the Black Coalition Against COVID, UK GP forums) documented numerous anecdotal cases of dark-skinned patients sent home from emergency departments with reassuring SpO2 readings who returned in extremis or died at home. The UK's NHS England COVID Oximetry @home programme — which distributed devices to millions of patients with the instruction to attend A&E if SpO2 dropped to 92% or below — became a focal point for advocacy after the publication of Sjoding 2020 and Fawzy 2022. Volume of reporting is high; the failure mode (delayed escalation, missed deterioration) was the outcome being reported, so survivorship bias is not the explanation.
Misconceptions
Three common ones to unlearn. First, that pulse oximeters work the same on everyone — the bias has been in the academic anesthesia literature for 35 years Jubran and Tobin 1990. Second, that race-based software correction is the fix; recent technical and ethical work explicitly rejects this approach — race is a proxy for skin pigmentation, not the variable itself, and embedding race into the algorithm relocates rather than solves the problem. The right fix is device-level validation against objectively measured skin tone (the direction of the FDA 2025 draft guidance) and bedside interpretation heuristics for existing in-service devices FDA 2025. Third, that the bias only matters at very low SaO2; mean error of 1–2 percentage points at SpO2 in the 92–96% range straddles the major clinical thresholds (88% as the floor for safe; 92% as the supplemental-oxygen trigger; 94% as the COVID-treatment gate), which is precisely why a small mean bias translates into large differences in care decisions.
Audience
The bias scales with skin pigmentation. Empirically, magnitudes have been demonstrated across Black, Hispanic, Asian, and South Asian populations, with the largest disparities in patients with the darkest pigmentation (Monk 8–10) Fawzy et al. 2022, Wong et al. 2021. Children show the same pattern. Light-pigmented patients are largely unaffected and can interpret SpO2 at face value. The article will write the interpretation heuristic for darker-pigmented readers and frame the rest as context for lighter-pigmented readers.
Failure-modes
Four predictable failure scenarios documented in the literature: (1) ward-based monitoring of acute respiratory illness, where 92% SpO2 is treated as adequate and supplemental O2 is undertitrated Gottlieb et al. 2022; (2) ED triage, where borderline SpO2 in an ambulatory patient is misread as "safe" and the patient is discharged or queued Fawzy et al. 2022; (3) at-home self-monitoring during COVID-style illness, where the layperson lacks the symptom-and-trend context a clinician would weigh, and the device is often an unregulated wellness product with bias on top of bias FDA 2021; (4) home sleep-apnea testing, where SpO2-derived desaturation indices are systematically lower in dark-pigmented patients, plausibly biasing diagnosis and severity classification — quantitatively under-characterized.
Practicalities
Most consumer fingertip oximeters sold in pharmacies and online (product codes PGJ and OCH) are classified as general-wellness devices, not medical devices; their accuracy is not reviewed by the FDA FDA 2021. Independent testing of low-cost wellness oximeters has shown wide variability and large absolute errors even in light-pigmented subjects, before the pigment-bias compounding. The first FDA-cleared OTC medical fingertip pulse oximeters (Masimo MightySat OTC, cleared 2024; Nonin TruO2) are designed and validated across skin tones and sit at $200–300. For clinical settings, the FDA 2025 draft guidance applies only to new device submissions; the millions of currently-in-service oximeters in hospitals are not retroactively re-validated.
History
Jubran and Tobin 1990 documented the issue in mechanically-ventilated patients; Bickler 2005 quantified it in a controlled hypoxia laboratory Jubran 1990, Bickler 2005. The follow-up by Feiner, Severinghaus, and Bickler in 2007 (Anesthesia & Analgesia) confirmed the gender-modulated pigmentation effect. The literature thus contains 35 years of warning. Conscious public-health attention and regulatory action arrived only after Sjoding 2020 and the COVID-19 pandemic's disparate mortality made the consequences visible at scale.
Out-of-scope
Adjacent topics the entry signposts but does not cover end to end: sleep apnea screening (where home pulse oximetry is the dominant test method and bias quietly propagates), wearable continuous SpO2 (Apple Watch, Oura, Garmin, Fitbit — wellness-class devices with mostly unstudied pigment bias), carbon monoxide poisoning (where standard pulse oximetry reads falsely normal at any pigmentation), and methemoglobinemia (similar artifact).
The credibility range
Optimist case
The mean bias magnitude is small — typically 1–2 percentage points Fawzy 2022. Experienced clinicians already triangulate SpO2 with symptoms, the trend, and ABG, so a single biased reading rarely makes a unilateral decision. The FDA is now actively tightening validation requirements FDA 2025, and two FDA-cleared OTC medical pulse oximeters that test across skin tones are now available. For the typical light-pigmented home user the device is reliable; for darker-pigmented users, the interpretation heuristic ("add ~2 to your action threshold") closes most of the practical gap without waiting for new hardware. The fix is in motion, even if the legacy installed base will be in service for a decade.
Skeptic case
The bias is structural, has been documented for 35 years, and was substantively ignored until pandemic-era excess mortality forced public attention. Mean bias of 1–2 percentage points understates the tail-risk distribution: occult hypoxemia rates 2–3× higher are not a small effect on the patients in that tail, and Wong 2021 ties hidden hypoxemia to a 41% mortality increase Wong 2021. The FDA 2025 guidance still relies partly on subjective skin-tone classification, applies only to new submissions, and excludes the consumer wellness oximeter market — i.e., the actual devices most home users own FDA 2025. Clinical heuristics ("subtract 2 if dark-pigmented") don't reliably propagate to ED physicians who didn't get the memo, and certainly don't propagate to the layperson reading a number on a $25 device during a COVID-style scare. The Hopkins COVID cohort estimate of hundreds of patients who never received qualifying treatments despite meeting the criterion by their actual blood gas Fawzy 2022 generalises across every pandemic respiratory illness and every hospital using pulse oximetry as a gating threshold.
Author's call
The phenomenon is settled science (evidence: high). The downstream care consequences are real and quantified (Fawzy, Gottlieb, Wong). The mean magnitude is small enough that for most readings well inside the normal range, the bias is not clinically actionable — but the disparity bites precisely at the borderline-but-okay-looking readings of 92–96% during acute respiratory illness, which is exactly the situation in which the device is most relied upon. The right reader takeaway: in any acute respiratory illness, symptoms and trend override device, and dark-pigmented readers should treat SpO2 92–93% as if it were 88–89%. For at-home use during a real concern, the FDA-cleared OTC medical devices are worth the premium over the drug-store wellness clip. Controversy on bias existence: low. Controversy on best fix: moderate — race-correction algorithms have been explicitly rejected, but engineering remedy vs clinical heuristic vs guidance enforcement remains an active argument.
Stakeholder and incentive map
- Device manufacturers (Masimo, Nonin, Nellcor/Medtronic, Philips): incumbent incentive to defend existing devices as adequate; newer entrants (Masimo MightySat OTC, Nonin TruO2) differentiate on skin-tone-tested accuracy.
- FDA: incentive to be visibly responsive post-COVID; 2022 and 2024 advisory panels, 2021 safety communication, 2025 draft guidance.
- Hospital systems and integrated payers (notably the VA via Iwashyna and Valbuena): purchasing leverage to drive market change by buying only validated devices.
- Patient advocacy and health-equity groups (Black Coalition Against COVID, UK race-equality NHS advocacy): pushing for transparency, mandatory on-device labeling, and regulation of wellness OTC devices.
- Academic critics (Lipnick et al., UCSF): the FDA 2025 guidance is necessary but insufficient — still permits subjective skin classification for half the cohort and excludes wellness OTC devices.
- Sleep medicine community: emerging concern that home sleep-apnea testing inherits the bias quietly.
Population variability
- Bias magnitude scales with skin pigmentation; largest in patients with the darkest measured pigment (Monk 8–10).
- Bias is larger at lower true SaO2 — i.e., grows in the zone where the reading actually matters Bickler 2005.
- Documented across Black, Hispanic, Asian, and South Asian patients Wong 2021, Fawzy 2022.
- Low peripheral perfusion (shock, hypothermia, vasopressors, peripheral vascular disease) compounds bias.
- Pediatric data shows similar pattern (Andrist et al., JAMA Network Open 2022 — referenced for completeness; not separately cited).
- Bias magnitude varies by oximeter model; same patient, different devices, different readings.
Knowledge gaps
- Wearable continuous SpO2 on consumer wrist devices (Apple Watch, Oura, Garmin, Fitbit) — pigment bias largely uncharacterized at scale; FDA 2025 guidance does not apply.
- Home sleep-apnea test accuracy (AHI/ODI by pulse oximetry) across pigmentation — under-quantified.
- Whether multi-wavelength reflectance pulse oximetry, or alternative chromophore-tolerant algorithms, can engineer the bias out at the device level.
- Whether the FDA 2025 cohort size (≥150 subjects, ≥25% dark) is statistically adequate to detect clinically meaningful tail-risk bias.
- Whether on-device labeling and standardized performance disclosure (Lipnick et al. proposal) changes user or clinician behavior.
- How fast the installed legacy device base will turn over, and whether retrofit calibration or software updates are feasible.
Scope vs brief. The brief named three consequences: occult hypoxemia detection, care escalation, and at-home use. All three covered end-to-end (mechanism + evidence + failure-modes sections handle the first two; practicalities + the at-home failure-mode bullet handle the third).
Dimension scoring. The entry sits awkwardly on the rating ladder because it is a "know" about a measurement bias, not an intervention. Most benefit dimensions score 0 honestly: the substance is awareness, and awareness only affects health/longevity through the rare averted-disaster case. Settled on health_short_term: 2 and longevity: 1 — both anchored on tail-risk cases (acute respiratory illness where the device's reading is the gating decision). Did not inflate to 3 to dodge the "thin benefit" optics; thin benefit is the honest read for the typical reader.
Audience scoping. Did not set audience.gender or audience.ages because the entry meaningfully applies to everyone — lighter-skinned readers as caregivers, darker-skinned readers as the directly-affected group. Folded the audience-differentiation into the protocol prose rather than a separate audience block.
Race vs pigmentation framing. The literature mostly uses self-reported race (because that's what is recorded in EHRs). The actual variable is skin tone. The article uses "darker skin / dark-skinned" rather than racial categories in reader prose; the research dossier preserves the original study language. Worth flagging because it can read as squeamish either way.
Numerical heuristic ("subtract two"). No formal guideline endorses this rule; it is the bedside rule of thumb most critical-care physicians familiar with the bias use, consistent with Fawzy 2022's 1.1–1.7 mean overestimate. Erring toward an actionable heuristic over hedged prose, but the rule is approximate, not authoritative. Flag for a clinician reviewer.
Excluded. Pediatric pulse-oximetry bias (Andrist 2022, JAMA Netw Open) — same direction, separate population. Sleep-apnea home oximetry bias — under-studied, separate entry candidate. Wearable wrist SpO2 (Apple Watch / Oura / Garmin / Fitbit) — wellness-class with mostly uncharacterized bias; signposted in out-of-scope.
Future-link candidates. Sleep apnea screening, carbon monoxide poisoning, COVID-19 home monitoring, ABG interpretation. None currently exist in the catalogue.
Separate-entry candidates. Sleep apnea screening (likely high-priority); CO poisoning (a "know" entry, useful for any home with combustion appliances); pulse oximetry's other failure modes (motion artifact, nail polish, methemoglobinemia) — could land as one general "limitations of pulse oximetry" entry or be folded into a parent "home medical monitoring" entry.
Hard call on tagline. Considered a more activist tagline ("Pulse oximeters miss low oxygen three times more often on darker skin") that leans on the Sjoding 2020 effect size. Chose the more practical "A 'green' 93% can hide a real 88%" because it tells the reader what to do with the number, not just that there's a problem.
Pulse Oximeter Pigment Bias
Knowing this is free. A device upgrade to a pulse oximeter tested across skin tones is optional and runs around $200–300.
Documented in hypoxia labs since the 1990s and confirmed across hospital cohorts of hundreds of thousands of readings. Recognised by the FDA.
Knowing how the bias works changes what you do with a borderline reading during a chest cold or COVID — the difference between sitting at home and getting checked.
Real but rare: it matters only in the tail case where a missed low-oxygen reading would otherwise be dismissed. In that case it can change the outcome.