Work to dismantle racism in breath testing won’t be as simple as trying to remove race from an algorithm, according to a new study from University of British Columbia researchers.
Breath tests are commonly used to diagnose or monitor asthma or other chronic lung diseases like chronic obstructive pulmonary disease and are done by blowing into a tube so a machine can measure how much and how fast a person can exhale, says Amin Adibi, a PhD candidate in the faculty of pharmaceutical sciences and the Respiratory Evaluation Sciences Program at UBC, and a UBC public scholar.
The problem is that breath testing has more than a century of racist assumptions baked into it because it was originally used in the U.S. to argue Black people have inferior lungs, Adibi says.
In 2023, leading American and European respiratory societies tried to implement a race-neutral approach to breath testing, but this latest study, led by Adibi and published in American Journal of Respiratory and Critical Care Medicine, highlights that no breath test can truly be race-neutral.
Therefore anyone administering a breath test needs to think critically about what test will create the most equitable findings for that situation, Adibi says.
“The problem is that it’s not like blood pressure,” Adibi said. “You cannot easily compare numbers between people because, for example, if someone is taller, you’re going to have bigger lungs and be able to exhale more.”
Lung capacity is generally linked to body size, he added.
To account for differences in body size the test results are compared to “reference values,” which calculate what the lung’s output should be based on a person’s biological sex, age, height and, until very recently, race and ethnicity.
“This test has a disturbingly racist history that goes back about 240 years all the way to Thomas Jefferson, the founding father and third president of the U.S.,” Adibi said.
Breath testing was invented in the United Kingdom to measure lung capacity based on occupation, and was picked up in the U.S. by Samual Cartwright, a pro-slavery plantation physician.
“He reported the lung function results were slightly lower in the Black population in the U.S. compared to the white population and this observation was weaponized by pro-slavery groups and, quite frankly, to justify white supremacy,” Adibi said. “They try to argue this is a sign that Black bodies are somehow inferior.”
Jefferson then wrote about these alleged differences in his book The State of Virginia, which may have helped the idea stick around.
In the 20th century industry used breath testing to measure fitness before hiring someone and Black people were often turned away because their test results would be slightly lower than white people’s, Adibi says.
Underdiagnosing, undertreating
Today, there is a better understanding of how social determinants of health impact people’s lives, for example how Black people were more likely to live in poorer, racialized neighbourhoods and be exposed to higher levels of pollution, Adibi said.
Biases in breath testing also mean that Black people have been underdiagnosed and undertreated for lung-related health issues because lower test results were seen as “benign” differences between white and Black bodies.
“Race is a social construct. It doesn’t have a straightforward biological meaning and it’s usually a proxy for social determinants of health which are very real and affect health,” Adibi said.
Ironically tests that included racial data were used because scientists felt it was a “thoughtful” way to incorporate people’s ethnicity, said Dr. Chris Carlsten, a professor and head of respiratory medicine at UBC, and director of Legacy for Airway Health.
Carlsten has worked with Adibi before but was not involved in the study.
Carlsten says that asking people about their race while administering breath testing is not particularly helpful because it can contribute to mistrust from the patient and add inaccuracy because racial categories are very imprecise.
Leading Black scientists W.E.B. Du Bois and Kelly Miller were pointing out how differences in lung capacity were due to social determinants of health, such as racialized communities being more exposed to pollution, more than 100 years ago, Adibi said.
It took the killing of George Floyd in 2020 to encourage a “critical mass” of scientists to re-evaluate race-based practices in medicine, Adibi said.
In 2023 the American Thoracic Society and European Respiratory Society jointly formed a workshop and recommended moving to a “race-neutral” equation for lung values.
Before this, patients were to self-identify as white, Black, northeast Asian, southeast Asian or “other,” which was an average of the other four categories, and then clinicians would look at their test results and compare them to what they “should” be. These categories excluded Indigenous identities.
The race-neutral equation takes a weighted average of the race-specific values.
This is where Adibi’s study comes in.
Defining ‘race-neutral,’ finding a way forward
As the definition of “race neutral” wasn’t clear or specific, Adibi’s study defined it in three ways.
The first is when patients don’t need to identify their race. The second is when test results will help diagnose the same ailment regardless of race. And the third is making sure the data doesn’t capture or encode race in some way.
“We showed that whether you use the old race-specific equation or the new race-average equations, there is no metric for lung function that would simultaneously meet all these definitions of race neutrality,” he said.
Race-specific lung testing didn’t meet the first definition, because patients have to self-identify, or the second, because the test outcomes meant different things for different people, but it did meet the third definition because the final test outcome is blind to race, Adibi said.
The new race-average way of testing does meet the first and second definitions but not the third, Adibi added. Patients don’t need to identify their race and the test results will help diagnose the same ailments regardless of race, but it fails the final definition because the test does not account for ongoing disparities in the social determinants of health.
This means people should be thinking critically about what equation they use in a breath test.
Adibi says if a clinician is using a breath test to assess patient risk for a longer surgery, then the race-averaged way of testing would be better.
But if you’re doing employment screening for new firefighters or determining life insurance premiums, maybe race-specific breath testing would be best, he says.
“The context is really, really important,” Adibi said.
Adibi’s study adds a puzzle piece to broader questions around algorithmic fairness and the concept of “the impossibility theorem.” In short: to be “fair,” you have to pick and choose your fairness metric depending on the context. There is no universal fairness.
Algorithms can exacerbate or mitigate racial disparities in health regardless of whether they include race or not, Adibi says. “It really doesn’t actually matter if the algorithm includes race or not, but how it handles it.”
“There is no simple solution to make sure an algorithm is not biased.”
Adibi’s study is important because race neutrality hasn’t been defined in this way before, Carlsten said.
It also highlights how race is associated with social determinants of health, and that’s important to look at because these are issues that can be improved or corrected with societal action or government or social policy, he said.