December 20, 2020

Racial Bias with Pulse Oximetry?

Background: Despite continued focus on equity in healthcare, it has yet to be realized in the United States. Healthcare can be considered a microcosm of society where resources are not allocated fairly among the population.  A recent editorial in the NEJM demonstrates that even equipment designed to give objective information can harbor important biases.  Pulse oximetry is a diagnostic tool that is ubiquitous in hospitals.  Triage and oxygen therapy administration decisions are made based on the readings from pulse oximetry. The development and validation of this tool was not performed in racially diverse populations but how this impacts the obtained information is unknown.

Paper: Sjoding MW et al. Racial Bias in Pulse Oximetry Measurement. NEJM 2020. PMID: 33326721

Clinical Question: How well do pulse oximetry readings correlate with arterial oxygen saturation readings on blood gas across white and black patients?

What They Did:

  • Analyzed paired pulse oximetry readings of oxygen saturation and arterial oxygen saturation from ABGs
  • All evaluations were performed within 10 minutes of each other
  • Limited analyses to measures of ABGs that included carboxyhemoglobin and methemoglobin

Outcomes: Occult hypoxemia (ABG O2 sat <88% despite an O2 sat of 92 to 96% on pulse oximetry)


  • Adult inpatients receiving supplemental oxygen at the University of Michigan and patients in the ICU wards at 178 hospitals


  • University of Michigan Cohort
    • Analyzed 10,789 pairs of oxygen saturation by pulse oximetry and arterial oxygen saturation in ABGs
    • 1333 white patients
    • 276 black patients
    • Occult Hypoxemia:
      • White patients: 99/2778 (3.6%; 95% CI 2.7 to 4.7)
      • Black patients: 88/749 (11.7%; 95% CI 8.5 to 16.0)
    • Multicenter Cohort
      • Analyzed 37,308 pairs of oxygen saturation by pulse oximetry and arterial oxygen saturations in ABGs
      • 7342 white patients
      • 1050 black patients
      • Occult Hypoxemia:
        • White patients: 546/8795 (6.2%; 95% CI 5.4 to 7.1)
        • Black patients: 160/939 (17.0%; 95% CI 12.2 to 23.3)


  • Peripheral perfusion could lower accuracy of oxygen saturation values, so the authors adjusted for age, sex, cardiovascular score on SOFA score in the University of Michigan cohort
  • Black patients had ≈3x the amount of occult hypoxemia not detected by pulse oximetry compared to white patients
  • The authors appropriately note that not all black patient who have a pulse oximetry value of 92 to 95% have occult hypoxemia
  • Pulse oximetry technology, and more broadly, medications and any medical technology, must be tested in multiple racial/ethnic groups is critically important in developing medical technologies
  • IMPORTANT TO NOTE: University of Michigan cohort all on supplemental oxygen and the multicenter cohort were all in the ICU which could limit generalizability to the ED
  • FOOD FOR THOUGHT: We don’t want to do unnecessary ABGs on patients, but we also don’t want to miss occult hypoxemia. Looking at the figure from above, when pulse oximetry reads 92 to 93% or less in black patients consider either watching patients in the hospital for observation (instead of discharging home) OR shared decision-making strategy in getting ABGs to ensure we are not missing occult hypoxemia (All patients will not need ABGs as this is a painful procedure and does have some risk associated with it).  Additionally as brought up in the discussion below, looking at the figure we can see that pulse oximetry has a +2% higher reading in black patients vs white patients across all readings.
  • ADDENDUM (12/21/2020): More thoughts on this topic as I have had discussions online with others…
    • I am unclear which pulse oximeters were being used in this study
    • Across the board there was a +2% higher reading in dark skin compared to light skin patients
    • This won’t impact clinical care until we get down to 92 – 93% or lower readings

Author Conclusion: “Our findings highlight an ongoing need to understand and correct racial bias in pulse oximetry and other forms of medical technology.”

Clinical Take Home Point: Pulse oximeters are used every day, but pulse oximetry may be less reliable in people of color by overestimating oxygen saturation in black patients when compared to white patients, especially when pulse oximetry reads 92 to 93% or less in black patients.


  1. Sjoding MW et al. Racial Bias in Pulse Oximetry Measurement. NEJM 2020. PMID: 33326721 

For More Thoughts on This Checkout:

Post Peer Reviewed By: Anand Swaminathan, MD (Twitter: @EMSwami)

Cite this article as: Salim Rezaie, "Racial Bias with Pulse Oximetry?", REBEL EM blog, December 20, 2020. Available at:
The following two tabs change content below.

Salim Rezaie

Emergency Physician at Greater San Antonio Emergency Physicians (GSEP)
Creator & Founder of REBEL EM
  • Albert Donnay
    Posted at 23:15h, 20 December Reply

    Thanks for reporting on this racial bias, but note the source is only a letter, not a NEJM editorial, and it needs editorial attention: The numbers of measurements reported in the text do not match the table (749 vs 713 Blacks, and 2778 v 2632Whites), and that the SpO2 ranges also do not match (92-96 in text vs 89-96 in table). Some correction is clearly needed and this may or may not lead to a retraction depending on whether any misconduct is involved (always a concern when data in text and figures don’t match!).

    But assuming one or the other are correct, the most important clinical take home point from these results is not that “pulse oximetry may be less reliable in people of color” but that it IS less reliable, and not just when displaying SpO2 in the 92-93% range. Although not mentioned by the authors, their figure clearly shows that SpO2 in Blacks is consistently 3% higher than the median arterial O2Hb by ABG, while this bias among Whites is only 0 to 1%.

    Until pulse oximeters come with skin tone settings, their manuals should recommend and clinicians should be taught to subtract 3% from ALL the SpO2 values of dark skinned people.

    • Salim Rezaie
      Posted at 07:17h, 21 December Reply

      Hello Albert,
      Will tackle these one at a time…

      1. Correct the numbers don’t exactly match, and that is a great catch by you. This is an important thing that needs to be followed up and clarified
      2. Agree that either way, it appears pulse oximeters are not calibrated for darker skin tones. The 92 to 93% or less I mentioned is for occult hypoxemia. Although it is true you could just subtract 2 or 3, this will not impact clinical care at the higher end of readings. It will however impact care as we get down to 92 to 93% or lower because as we subtract three we fall into occult hypoxemia range

      Appreciate you reading and your insightful comment.


  • Michael Sjoding
    Posted at 13:26h, 21 December Reply

    Thanks for catching this! We have identified the source of the error (which is in the figure not the main text) and will ask NEJM for a correction,

    • Salim Rezaie
      Posted at 13:43h, 21 December Reply

      TY Michael…appreciate your hard work and quick response. Great piece and one we should all be aware of.


  • Albert Donnay
    Posted at 15:12h, 22 December Reply

    Glad Dr. Sjoding found the problem and will request correction. Can he tell us if this will change the figure or just the numbers of Black and White underneath? And if he looked, what was the range of bias in the Black vs White measures when SpO2 displayed 97-99%? Extrapolating from the original figure, it looks like the B measures are headed for +5% bias at SpCO 99% and W for +3%, which would be consistent with manufacturers trying to keep their (mostly White) calibration studies within the FDA accuracy limit of +/-3%.

Post A Comment

Time limit is exhausted. Please reload CAPTCHA.