Pulse oximeters became a central tool in managing COVID-19 during the pandemic, particularly through community monitoring schemes such as the NHS England COVID Oximetry @home programme. These small fingertip devices allowed patients to monitor peripheral oxygen saturation at home and seek care early if levels dropped. However, concerns have grown about whether pulse oximeters perform equally well across different skin tones. Inaccurate readings can delay the diagnosis of hypoxaemia and contribute to avoidable harm.
A major measurement and diagnostic accuracy study published in The BMJ in January 2026 provides the most comprehensive evidence to date on how skin tone affects the performance of pulse oximeters used in the NHS England COVID Oximetry @home scheme. Known as the EXAKT study, this research used objective spectrophotometry rather than ethnicity or race as a proxy for skin tone. The findings have important implications for patient safety, clinical guidance, and medical device regulation.
This article reviews the key findings of the study, explains why they matter, and explores their implications for clinical practice and future research.
Pulse oximetry estimates arterial oxygen saturation by transmitting red and infrared light through tissue and measuring absorbance differences between oxygenated and deoxygenated haemoglobin. While this technology has been in use since the 1980s, it was originally validated in relatively small groups of healthy volunteers, many of whom had lighter skin tones.
Previous research has shown that pulse oximeters may overestimate oxygen saturation in people with darker skin. This can lead to occult hypoxaemia, a situation where arterial oxygen levels are dangerously low but the pulse oximeter reading appears reassuring. Many earlier studies relied on ethnicity or race as a surrogate for skin tone or used subjective colour charts, which limits precision and interpretability.
The EXAKT study was designed to overcome these limitations by directly measuring skin tone using spectrophotometry and by evaluating pulse oximeters in a large cohort of critically ill patients.
The EXAKT study was a prospective measurement and diagnostic accuracy study conducted in 24 NHS intensive care units in England between June 2022 and August 2024. A total of 903 adult patients receiving mechanical ventilation and supplemental oxygen were enrolled.
Five fingertip pulse oximeters supplied by NHS England for the COVID Oximetry @home scheme were tested:
These devices represent the type of low cost, battery powered oximeters used by patients at home during the pandemic.
Pulse oximeter readings of peripheral oxygen saturation, or SpO2, were compared with arterial oxygen saturation, or SaO2, measured using blood gas co-oximetry. SaO2 is considered the gold standard for assessing oxygenation.
Over 11,000 paired SpO2 and SaO2 measurements were collected. Skin tone was objectively assessed using a handheld spectrophotometer and expressed as the individual typology angle, or ITA. This provided a continuous and reproducible measure of skin tone ranging from very light to dark.
The study examined both measurement accuracy and diagnostic accuracy:
Two clinically relevant SpO2 thresholds were analysed: 92 percent and 94 percent. These thresholds align with NHS guidance for escalation of care in the COVID Oximetry @home scheme.
Across all devices, pulse oximeters tended to overestimate oxygen saturation at lower SaO2 levels and underestimate it at higher levels. There was substantial variability between devices, with no single oximeter consistently outperforming the others across all measures.
For any given SaO2, all five pulse oximeters produced higher SpO2 readings in patients with darker skin tones compared with those with lighter skin tones. On average, SpO2 readings were between 0.6 and 1.5 percentage points higher in patients with dark skin tone, defined as an ITA around minus 44 degrees, compared with those with light skin tone, defined as an ITA around 46 degrees.
Although this difference may appear small, it had important downstream effects on diagnostic accuracy.
False negative rates for detecting hypoxaemia increased substantially with darker skin tone. At an SpO2 threshold of 94 percent, the proportion of readings that failed to identify true hypoxaemia was between 5.3 and 35.3 percentage points higher in patients with darker skin compared with those with lighter skin. In relative terms, false negative rates were up to seven times higher.
In contrast, false positive rates decreased with darker skin tone. This means that patients with lighter skin were more likely to be incorrectly identified as hypoxaemic, while patients with darker skin were more likely to have hypoxaemia missed.
Occult hypoxaemia, defined as SaO2 below 88 percent with SpO2 above 92 percent, was more common in patients with darker skin tones. Although the number of such cases was small, the trend was consistent across devices and aligns with concerns raised in earlier studies.
The EXAKT study demonstrates that skin tone systematically affects pulse oximeter readings in a clinically meaningful way. Importantly, the study challenges the assumption that overall accuracy metrics such as ARMS are sufficient to ensure equitable device performance. ARMS was largely driven by imprecision and failed to capture differences in bias that translated into large disparities in false negative rates.
The findings also show that the effect of skin tone is not simply a matter of worse performance in darker skin across all conditions. At higher oxygen saturations, higher SpO2 readings in darker skin sometimes reduced underestimation. However, at clinically critical lower saturations, the same bias increased the risk of missing hypoxaemia.
Clinicians and patients should be aware that pulse oximeter readings are estimates rather than definitive measurements. This is particularly important when monitoring individuals with darker skin tones.
Key practical implications include:
The current international standard for pulse oximeter performance requires testing in as few as ten individuals and does not mandate representation across skin tones. The EXAKT study shows that such standards are insufficient to detect clinically important bias.
Recent draft guidance from the US Food and Drug Administration recommends larger and more diverse validation studies, including objective assessment of skin tone. However, the thresholds proposed for acceptable variation in bias may still allow differences that produce substantial disparities in diagnostic accuracy.
Regulators and manufacturers should consider incorporating more stringent criteria focused on diagnostic outcomes rather than relying solely on aggregate accuracy metrics.
Further research is needed to evaluate pulse oximeter performance in community settings using objective skin tone measures. Studies should also assess higher fidelity hospital devices and explore technical approaches to reducing skin tone related bias.
In addition, there is a need to identify practical alternatives to spectrophotometry for assessing skin tone in research and regulatory testing, particularly in low resource settings.
The EXAKT study provides robust evidence that fingertip pulse oximeters used in the NHS England COVID Oximetry @home scheme yield higher SpO2 readings in patients with darker skin tones. Although the absolute differences in measurement were modest, they translated into large disparities in the ability to detect hypoxaemia.
These findings highlight the need for more inclusive device validation, improved clinical guidance, and greater awareness of the limitations of pulse oximetry. Addressing these issues is essential to ensure equitable and safe care for all patients.
Martin D S, Doidge J C, Gould D, Shahid T, Cowden A, Charles W N et al. The impact of skin tone on performance of pulse oximeters used by NHS England COVID Oximetry @home scheme: measurement and diagnostic accuracy study BMJ 2026; 392 :e085535 doi:10.1136/bmj-2025-085535
This article is based on research published in The BMJ and is distributed under the Creative Commons Attribution (CC BY) Open Access licence. The interpretation and summary presented here are for informational and educational purposes only and do not constitute medical advice. Readers should consult qualified healthcare professionals for clinical decision making.
\
Most Accurate Healthcare AI designed for everything from admin workflows to clinical decision support.