Pulse oximeters are less accurate for Black people

Pulse oximeters are ubiquitous, noninvasive tools for measuring blood oxygen levels in medical settings. They clip onto a finger, a toe, or an earlobe and calculate the oxygen level in a person’s blood by shining different wavelengths of light through the skin. But a recent study reports that these devices often overestimate oxygen levels in Black people. Consequently, they fail to detect critically low levels of oxygen, a condition called hypoxemia, approximately three times as often in Black people as in white people (N. Engl. J. Med. 2020, DOI: 10.1056/NEJMc2029240).

The findings are especially alarming during the COVID-19 pandemic, says Michael Sjoding, a pulmonary and critical care physician at the University of Michigan Medical School and a lead researcher on the project. COVID-19 has killed a disproportionate number of Black people and other people of color in the US. Hypoxemia is a key feature of COVID-19, and health-care providers frequently use pulse oximetry readings to determine whether a patient should be hospitalized or receive supplemental oxygen.

Pulse oximeters emit light at two wavelengths—one red and one infrared. Hemoglobin, the protein in red blood cells that carries oxygen, occurs in two forms, and each absorbs different wavelengths of light. The oxygen-bound form, oxyhemoglobin, absorbs more infrared light; the oxygen-free form, deoxyhemoglobin, absorbs more red light. Pulse oximeter sensors use the relative ratio of how much of each wavelength of light gets absorbed to calculate oxygen levels.

Sjoding and his colleagues launched the study because they noticed in the early stages of the pandemic that Black people being treated for COVID-19 often showed a normal pulse oximetry reading but low oxygen levels on an arterial blood gas test, which requires a blood sample.

They systematically compared pulse oximetry and arterial blood gas measurements in people who received supplemental oxygen from January to July of 2020, many of whom had COVID-19, at the University of Michigan Hospital. They included people who listed their race as Black and people who listed their race as white. The team looked for cases of hypoxemia that were not flagged by pulse oximetry. In these cases, arterial oxygen saturation was below 88%, a potentially dangerous level, but pulse oximetry registered values of 92–96%, within the range of normal. The team found that 11.4% of Black patients showed this pattern, compared with 3.6% of white patients. The researchers found similar discrepancies in data from people treated in the intensive care units of multiple hospitals from 2014 to 2015.

Sjoding notes that the study did not look at skin color directly. “We didn’t know the skin color of people in our data—just whether they self-reported Black or white,” he says. The findings likely apply not just to Black people but more generally to people with dark skin, Sjoding says, because the amount of pigment changes how light is transmitted through skin.

“We rely so much on oxygen saturation on a routine basis that this is disconcerting,” says Taison Bell, a critical care physician at UVA Health at the University of Virginia who was not involved in the work. “Those oxygen levels correlate with low oxygen that goes to the brain and other vital organs,” he says. “It makes you wonder, are we putting ourselves way behind the game [with Black patients] by depriving their vital organs of oxygen, compared to white patients?”

Studies over the past few decades have reported discrepancies in pulse oximetry readings based on skin color. But that older work generally identified the discrepancies in cases in which oxygen-saturation levels were already low enough that patients would likely already be receiving treatment, Sjoding says. What’s more, those results were not widely disseminated. Even among pulmonary experts, “the fact that there’s bias in this technology was just not well known,” he says.

Yasser Khan, an electrical engineer at Stanford University who develops medical devices, says the problem is overlooked not only among clinicians but also among medical device makers. The mathematical calculation that converts the light measurements into a value for oxygen saturation needs to be calibrated to account for the uneven way that skin, bones, blood, and tissue disperse the red light. The calibration depends on skin color because darker skin disperses more light than lighter skin, Khan explains. So if people with dark skin are underrepresented in the calibration process, he says, “the calibration curve won’t work for them and there will be a bias in the interpreted oxygenation value.”

As a start, Khan says, pulse oximeters must be tested on a more diverse group of people, but fixing the bias is an engineering problem to be addressed at the level of the sensor, the hardware, the software, and the calibration. Sjoding says that since the paper’s publication, two device manufacturers have contacted him saying that they may have a solution to the problem.

Bell cautions that researchers and device makers will need to take care not to use race as a proxy for skin color but to correlate pulse oximetry levels with actual skin tones. “It would be too blunt an instrument” to apply one scale for Black people and another for white people, since that overlooks other people of color and since skin colors vary widely even among people who identify as the same race, he says.