For their study, published in The BMJ and funded by the National Institute for Health and Care Research (NIHR), the researchers drew on data from 903 critically ill adults (average age 56 years; 67% male) on 24 NHS intensive care units in England between June 2022 and August 2024.
Although the fingertip pulse oximeters tested were those used at home, intensive care units were used as the test laboratory as patients have lower blood oxygen values and routinely have their blood oxygen accurately measured using hospital machines.
For each patient, skin tone was measured objectively using a spectrophotometer (a type of camera that measures colour) and researchers then compared pulse oximetry blood oxygen values (SpO2) with “gold-standard” arterial blood gas measurements (SaO2). SpO2 values were assessed at two thresholds in line with current guidance either to seek medical help (94% or lower) or to attend the emergency department (92% or lower).
A total of 11,018 paired SpO2-SaO2 measurements were analysed. All five pulse oximeters returned higher SpO2 values for patients with darker skin tones than patients with lighter skin tones, at any given level of SaO2 measurement.
SpO2 readings were, on average, 0.6-1.5 percentage points higher for patients with darker than lighter skin tone.
At both SpO2 thresholds assessed, false negative rates (low oxygen missed by the pulse oximeter when present) increased with darker skin tones, while false positive rates (low oxygen indicated by the pulse oximeter when absent) decreased with darker skin tone.
And while absolute differences in readings were small, the researchers say “they can result in substantially higher rates of false negatives and lower rates of false positives in the diagnosis of hypoxaemia”.
The observational nature of the study means no definitive conclusions can be drawn about cause and effect, and the authors acknowledge that the study was conducted in critically ill patients and so may limit the generalisability of the findings.
However, they say this was a large study using sophisticated statistical modelling to assess the performance of the pulse oximeters across multiple domains of both measurement and diagnostic accuracy.