One and a half years ago, a landmark study by Sjoding et al. found that critically ill Black patients were more likely than White patients to have falsely reassuring pulse oximetry measurements (discussed here).1 Subsequently, six sizable studies have been published on this topic. Since this is of critical importance to acute care medicine, it's worth reviewing the new data to see where we stand.
Wong et al. 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, 2021.
This was a retrospective analysis of several public ICU databases (e.g. the MIMIC database from Beth Israel Deaconess hospital and databases from Grady Memorial Hospital).2 87,971 patients were identified with a combination of a saturation >88% and a corresponding arterial blood gas. Hidden hypoxemia was defined as an oxygen saturation on ABG of <88% despite a pulse oxygen saturation >88%.
The rate of hidden hypoxemia was higher among Black patients (6.9%) and Hispanic patients (6.0%) as compared to White patients (4.9%) or Asian patients (4.9%) (p<0.001). This supports the presence of bias.
Burnett GW et al. Self-reported race/ethnicity and intraoperative occult hypoxemia: A retrospective cohort study. Anesthesiology, 2022.
This is an analysis of a database from Mount Sinai School of Medicine including 46,523 patients who underwent anesthesia.3 Pulse oximetry was compared with oxygen saturation as calculated based on PaO2 measured using a point-of-care ABG analyzer.4
Occult hypoxemia was defined as a saturation <88% based on the ABG analyzer, despite pulse oximetry >92%. Occult hypoxemia was more prevalent among Black and Hispanic patients (table below). Even in multivariate analysis, these correlations remained significant.
The inaccuracy of pulse oximetry (mean +/- standard deviation) was -0.2% +/- 6.3% in White patients, +0.6% +/- 9.1% in Black patients, and +0.5% +/- 7.9% in Hispanic patients. Thus, a greater rate of occult hypoxemia in Black and Hispanic patients was due to positive average inaccuracy (+0.7-0.8% relative to White patients) combined with a higher standard deviation (i.e., reduced accuracy and reduced precision).
Nellcor pulse oximeters were used prior to 2011 and Masimo were used thereafter. There were no obvious differences in the rates of occult hypoxemia after changing from Nellcor to Masimo pulse oximeters.
Crooks CJ et al. Pulse oximeters' measurements vary across ethnic groups: An observational study in patients with Covid-19 infection. European Respiratory Journal, 2022.
This study investigated an electronic database from Nottingham University Hospitals NHS trust to compare pulse oximetry measurements with ABG obtained within a 30-minute time window.5 Pulse oximetry was aberrantly elevated in Black and Asian patients:
Due to the relatively long time intervals between ABG and pulse oximetry, precise differences may not be accurate. However, relative differences in accuracy should remain reliable.
Henry NR et al. Disparities in hypoxemia detected by pulse oximetry across self-identified racial groups and associations with clinical outcomes. Critical Care Medicine, 2022.
This is an analysis of a database of patients at three Mayo Clinic locations who were undergoing surgery or were admitted to the intensive care unit.6 Pulse oximetry was compared to simultaneous arterial blood gas values. Occult hypoxemia was defined as an oxygen saturation <88% on arterial blood gas analysis despite a pulse oximetry saturation of 92% or higher. The risk of occult hypoxemia was higher in all non-White groups:
Valbuena VSM et al. Racial bias in pulse oximetry measurement among patients about to undergo extracorporeal membrane oxygenation in 2019-2020. Chest, 2022.
This is an analysis of an ECMO database including data from 324 centers.7 Pulse oximetry was compared with ABG values during six hours prior to initiation of ECMO. Occult hypoxemia was more likely among Black patients:
Black patients had a higher mean bias (+1.7%) as well as a wider standard of deviation, as compared to White patients (table below). Hispanic patients had an intermediate mean bias:
Fawzy A et al. Racial and ethnic discrepancy in pulse oximetry and delayed identification of treatment eligibility among patients with COVID-19. JAMA Internal Medicine, 2022.
This is a retrospective analysis of COVID patients within five hospitals in the Johns Hopkins Health system between 3/20-12/21.8 Pulse oximetry was compared with ABG oxygen saturation (measured using co-oximetry) within ten minutes.
Pulse oximetry was found to overestimate the actual arterial oxygen saturation among Asian, Black, and non-Black Hispanic patients:
Occult hypoxemia was defined as a true oxygen saturation <88% despite a pulse oxygen saturation measurement of 92-96%. Consistent with the table above, the rate of occult hypoxemia at some point during hospitalization was higher for Asian (30%), Black (29%), and non-Black Hispanic (30%) patients, as compared to White patients (17%).
Overview of new data
Prior studies dating back three decades have shown that pulse oximetry may be inaccurate among Black patients.9 New studies confirm that this remains a significant and widespread problem. The magnitude of the average error isn't enormous (e.g., 1-2%). However, when combined with impaired precision, this is large enough to commonly cause clinical harm due to undertreatment of hypoxemic patients.
Several of the above studies found associations between occult hypoxemia and worse clinical outcomes. I haven't focused on these associations, since it's impossible to differentiate causation from correlation. However, it should be intuitively clear that impaired detection of illness severity will lead to inadequate treatment and worse outcomes.
The problem doesn't appear to be restricted to Black patients, either. Studies have detected impaired precision in a variety of racial groups, including Hispanic, Asian, and American Indian patients. Results are somewhat variable between studies, possibly reflecting that these populations are not homogeneous (e.g., greater skin pigmentation is often present in people of Southern Asian descent compared to those from Northern Asia).10 Nonetheless, it's highly likely that pulse oximetry dysfunction extends well beyond Black patients.
These studies do have important limitations, most notably:
- It's frequently unclear precisely which pulse oximeters were used. As explored in a prior post, different models may perform differently. This may be another source of variability across different studies.
- Race is a poor indicator of skin tone.10
- Many studies don't clarify whether oxygen saturation from ABG is measured via co-oximetry or calculated from a PaO2 value (which is common practice with point-of-care ABG monitors).
Pulse oximetry sorely needs an update
The fundamentals of pulse oximetry have hardly changed in the past thirty years. This is remarkable, when considering the evolution of other technological devices over the same time frame (e.g., computers, smartphones, CT scanners).
Clinical medicine has a tendency towards inertia. So we accept pulse oximetry as what it is now – cheap and pretty accurate, but often incorrect. Current pulse oximetry falls flat on its face among patients with carbon monoxide poisoning or methemoglobinemia – emergencies which a better device could easily detect. There is also room for improvement in the evaluation of patients with poor perfusion (which sometimes causes pulse oximeters to spit out obviously erroneous data). Clinicians are trained to anticipate and compensate for these limitations, deterring us from demanding a better device.
Pulse oximetry could benefit from a fundamental redesign to improve its performance across the board. For example, measuring more than two wavelengths would provide a degree of redundancy, allowing the device to determine if incoming data is internally consistent. This could allow the oximeter to warn clinicians if the data is unreliable (e.g., due to methemoglobin, carboxyhemoglobin, or poor perfusion). Additional wavelengths could also detect and correct for skin pigmentation.11 The best way to eliminate a medical device is to replace it with a better device.
Of course, there is a moral imperative to fix racial bias in pulse oximetry. This is required to achieve justice for groups which have been discriminated against by the medical establishment. Eliminating racial bias in pulse oximetry might also help restore faith in the medical establishment.
Unfortunately, it's dubious whether justice alone is a sufficient stimulus to effect actual change. The medical-industrial complex is historically intransigent in the face of inequities. Overall, progress in the past year and a half has been disappointing. The Food and Drug Administration (FDA) issued a warning about pulse oximetry over a year ago. A couple of medical societies have issued statements on the topic, but most have remained silent (kudos to the Intensive Care Society). There doesn't appear to be substantial grant funding nor legislation directed towards the issue. Nobody has performed a rigorous head-to-head trial comparing the performance of different oximeter brands among people of color. Every day that passes without action pushes us closer towards tacit acceptance and further inaction.
- Pulse oximetry overestimates oxygen saturation among patients with dark skin by an average of 1-2%. Coupled with impaired precision, this creates a clinically substantive risk of missing hypoxemia.
- In addition to Black patients, this problem probably affects anyone with skin tones darker than Caucasians (e.g., many Hispanic, Asian, and American Indian patients).
- The past year and a half have confirmed that racial bias in pulse oximetry is a real and widespread problem, but little action has yet been taken on this front.
- Dismantling the systemic racism of pulse oximetry (PulmCrit)
- Racial bias in pulse oximetry (Salim Rezaie in RebelEM)
- More on pulse oximetry and racial bias (Simon Carely at St Emlyns)
- Oximeters used to be designed for Equity. What happened? (Wired)
- 1.Sjoding M, Dickson R, Iwashyna T, Gay S, Valley T. Racial Bias in Pulse Oximetry Measurement. N Engl J Med. 2020;383(25):2477-2478. doi:10.1056/NEJMc2029240
- 2.Wong A, Charpignon M, Kim H, et al. Analysis of Discrepancies Between Pulse Oximetry and Arterial Oxygen Saturation Measurements by Race and Ethnicity and Association With Organ Dysfunction and Mortality. JAMA Netw Open. 2021;4(11):e2131674. doi:10.1001/jamanetworkopen.2021.31674
- 3.Burnett G, Stannard B, Wax D, et al. Self-reported Race/Ethnicity and Intraoperative Occult Hypoxemia: A Retrospective Cohort Study. Anesthesiology. 2022;136(5):688-696. doi:10.1097/ALN.0000000000004153
- 4.Bénéteau-Burnat B, Bocque M, Lorin A, Martin C, Vaubourdolle M. Evaluation of the blood gas analyzer Gem PREMIER 3000. Clin Chem Lab Med. 2004;42(1):96-101. doi:10.1515/CCLM.2004.018
- 5.Crooks C, West J, Morling J, et al. Pulse oximeter measurements vary across ethnic groups: an observational study in patients with COVID-19. Eur Respir J. 2022;59(4). doi:10.1183/13993003.03246-2021
- 6.Henry N, Hanson A, Schulte P, et al. Disparities in Hypoxemia Detection by Pulse Oximetry Across Self-Identified Racial Groups and Associations With Clinical Outcomes. Crit Care Med. 2022;50(2):204-211. doi:10.1097/CCM.0000000000005394
- 7.Valbuena V, Barbaro R, Claar D, et al. Racial Bias in Pulse Oximetry Measurement Among Patients About to Undergo Extracorporeal Membrane Oxygenation in 2019-2020: A Retrospective Cohort Study. Chest. 2022;161(4):971-978. doi:10.1016/j.chest.2021.09.025
- 8.Fawzy A, Wu T, Wang K, et al. Racial and Ethnic Discrepancy in Pulse Oximetry and Delayed Identification of Treatment Eligibility Among Patients With COVID-19. JAMA Intern Med. Published online May 31, 2022. doi:10.1001/jamainternmed.2022.1906
- 9.Jubran A, Tobin M. Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent patients. Chest. 1990;97(6):1420-1425. doi:10.1378/chest.97.6.1420
- 10.Norton H. Variation in pulse oximetry readings: melanin, not ethnicity, is the appropriate variable to use when investigating bias. Anaesthesia. 2022;77(3):354-355. doi:10.1111/anae.15620
- 11.Ferrari M, Quaresima V, Scholkmann F. Pulse oximetry, racial bias and statistical bias: further improvements of pulse oximetry are necessary. Ann Intensive Care. 2022;12(1):19. doi:10.1186/s13613-022-00992-z
- PulmCrit Hot Take: Aggressive glycemic control is dead (TGC-Fast Trial) - September 28, 2023
- PulmCrit – Introducing the IBPH (Internet Book of Hospital Pulmonology) - September 3, 2023
- PulmCrit: “ARDS” is not a real thing - May 27, 2023