Are Pulse Oximeters "Intrinsically Biased Against People of Color?"

A Wall Street Journal article suggests that pulse oximeters mismeasure blood oxygen levels in American minority populations, but the article fails to ask some important questions.

I am a huge advocate of using pulse oximeters, which measure blood oxygen saturation levels, to help monitor how Covid is affecting one's lungs. A pulse oximeter may have even saved my life; when I caught Covid in fall 2020, I had no trouble breathing but my handy pulse oximeter told me that my measurements were at a level where I should seek medical attention, and physicians in my family told me to go immediately to Georgetown Hospital. Once there, they found significant pneumonia in my lungs, admitted me, treated me, and released me good as new (after a few weeks recovery).

One concern about pulse oximeters, however, has been that they may work less well on people with dark skin. Specifically, that they slightly overestimate blood saturation levels; that slight overestimation, however, can have significant consequences regarding medical attention and treatment.

Today, the Wall Street Journal breathlessly reports on a study that seems to confirm that the concern is warranted.

The recent study looked at more than 7,000 patients diagnosed with Covid-19 between March 2020 and November 2021 at five hospitals in the Johns Hopkins Health System.

Pulse oximeter measurements were taken for all 7,000 patients. More-accurate but invasive arterial blood gas measurements also were taken from about 1,200 of the patients.

When the researchers compared measurements from the arterial blood gas tests with pulse oximeter readings, they found discrepancies among patients of all races but a more pronounced difference for patients of color. Compared with white patients, pulse oximeters overestimated oxygen saturation by an average of 1.2 percentage points among Black patients, 1.1 percentage points among Hispanic patients and 1.7 percentage points among Asian patients.

Some doctors, not quoted, expressed reservations to the Journal, but Dr. Martin Tobin stated that the study shows that pulse oximeters are "intrinsically biased against people of color." That's an oddly unscientific way of putting it; at most, we can say that the device gives less accurate results for people with darker skin pigmentation. Many "white" people have darker skin than many Hispanic, Asian, and some Black Americans.

Which brings me to some concerns about the study. First, you want to measure how accurate pulse oximeters are for people with dark skin, why not directly study that? Self-identified race and ethnicity is a crude substitute for skin tone.

This is particularly true given the internal diversity within each category. The study in question used data from the Johns Hopkins health system. Is the Hispanic and Asian population of the mid-Atlantic representative of the national Hispanic and Asian population in terms of skin tone, given that they have populations with different national origins (e.g., way fewer Mexican Americans on the East Coast than in the nation as a whole)? If so, how dark-complexioned are they compared to whites and African Americans? I doubt anyone has even considered those questions.

And then the results themselves are odd, if skin tone is really the decisive factor. Recall that the study finds that "pulse oximeters overestimated oxygen saturation by an average of 1.2 percentage points among Black patients, 1.1 percentage points among Hispanic patients and 1.7 percentage points among Asian patients." I don't need a scientific study to tell me that the average African American is darker-complexioned than the average Hispanic or Asian American (though if you want scientific studies, we do know from DNA studies that Hispanic Americans have a much higher percentage of European ancestry than do African Americans, which would be reflected in average skin tones.) If dark complexion were the issue, we would expect the overestimation to be greatest among Black patients, not Asians, and we would expect the gap between Black and Hispanic patients to be greater.

The reporter, apparently caught up in the "people of color" paradigm, does not seem to even notice these questions.

The article concludes:

The FDA recommends accuracy studies include 10 or more healthy subjects who vary in age and gender. At least two participants, or 15% of the participant pool, should be "darkly pigmented." The FDA said it was evaluating whether its guidance should be modified.

It strikes me that what needs to be done is a wide scale study that ignores America's unscientific official racial and ethnic categories and directly measures subjects' skin pigmentation, and measures whether darker skin pigmentation correlates with less accurate results. And if the government wants to measure whether pulse oximeters indeed work less well on, e.g., "Asian Americans" for reasons other than complexion, it needs to break the categories down by subgroup; there is no particular genetic or sociological commonality between South Asians like Indians and Pakistanis and East Asians like Chinese and Vietnamese.