Covid-19 care, including distribution of lifesaving therapies, was significantly delayed for Black and Hispanic patients due to inaccurate oxygen readings from devices that can work poorly in darker-skinned individuals, according to a study published Tuesday. The finding may be one reason much higher Covid-19 mortality rates have been seen in communities of color across the United States.
Widely used pulse oximeters, which measure oxygen levels by assessing the color of the blood, have been under increasing scrutiny for racial bias because they can overestimate blood oxygen levels in darker-skinned individuals and make them appear healthier than they actually are. A 2020 study comparing oxygen levels measured by the devices with readings taken from “gold standard” arterial blood samples found pulse oximeters were three times less likely to detect low oxygen levels in Black patients than in white patients. Two months after that report, the Food and Drug Administration issued a safety communication alerting patients and clinicians that the devices could be erroneous in those with dark skin.
The study, published in JAMA Internal Medicine, shows the inaccuracies in oxygen measurement occurred at higher rates than in white patients not only in Black patients, but also in Hispanic and Asian patients, and that those inaccuracies had real-world consequences. The study provided evidence that undetected low oxygen levels led to delays in Black and Hispanic patients receiving potentially lifesaving therapies such as the drugs remdesivir and dexamethasone, and in many cases, led to patients not receiving treatment at all.
“These are likely patients who were seen in emergency rooms and sent home,” said Tianshi David Wu, an assistant professor of medicine at Baylor College of Medicine and co-lead author of the new study. He called pulse oximeters “de facto gatekeepers” for Covid treatment because low readings on these devices are key criteria for deciding how aggressively patients should be treated. “There are patients that probably should have had these therapies, and the majority were Black patients.”
The study used data from more than 7,000 patients who were seen in the emergency departments or admitted at one of five hospitals in the Johns Hopkins Health System in Baltimore. It found that more than one-fourth of them — mostly people from marginalized racial and ethnic groups — would have qualified for Covid-19 therapies before the pulse oximeter recognized their need of care. They found Black patients were 29% less likely than white patients to have eligibility for treatment recognized by the devices while non-Black Hispanic patients were 23% less likely to have their eligibility for treatment recognized.
The authors, both pulmonary critical care physicians who have been treating Covid patients, said they were motivated to determine whether the inaccuracies in readings were clinically relevant because they felt many physicians — despite the fact that concerns about bias in pulse oximeters have been raised in the scientific literature for decades — remained unaware of how the inaccuracies in these critically important diagnostic devices might impact patients of color. The issue has not been part of standard medical school curricula, they said.
“It’s not like this is new info, but I can certainly say I was not instructed about this,” said Ashraf Fawzy, an assistant professor of medicine at Johns Hopkins University and a co-lead author of the study. “There is clinical relevance to this pulse oximeter inaccuracy, more so in racial and ethnic minorities. It’s a systematic failure.”
The study found that pulse oximeters overestimated oxygen levels by 1.2% in Black patients, by 1.1% in Hispanic patients, and by 1.7% in Asian patients. Those may sound like small differences, but because oxygen levels are so critical to health, treatment protocols for many diseases, including Covid, can differ dramatically when small differences are detected in oxygen levels.
For Covid, drug treatments are indicated when oxygen levels fall below 94%, yet the study showed many patients had blood oxygen levels below that level, despite what their pulse oximeters read. “It’s a good illustration of how a relatively small bias in accuracy can have a large effect,” said Wu.
Black patients whose eligibility for Covid treatment was eventually confirmed by pulse oximetry had treatment delayed by an hour, compared to white patients, the researchers found. (Patients who never received treatment were not included in this accounting.) Wu said it was not clear if that delay was clinically significant. The authors said more studies were ongoing to assess if such delays or lack of access to treatments led directly to greater mortality.
The study is an excellent addition to accumulating evidence that the inaccuracies in these devices are taking a large toll, said Thomas Valley, an assistant professor of medicine at the University of Michigan whose publication with colleagues in 2020 about the inaccuracy of devices in Black patients helped stoke widespread interest in how the bias may affect care. Previous studies have shown that undetected low oxygen rates can lead to sequential organ failure and death in patients of color, but those studies were not conducted specifically on Covid patients as the new study was.
“We’ve been searching for reasons Black and Hispanic people were more likely to die early in the pandemic,” Valley said. “This is pretty depressing that we had treatments available, many of those treatments tied to oxygen levels. Decisions on whether or not people were admitted to hospitals or put on ventilators, those were all based on blood oxygen levels.”
Valley said he’d like to see a study on how decisions to send people home from emergency departments or tell them not to come to the hospital based on blood oxygen levels may have impacted patients with darker skin. “We were recommending that all the time, ‘If your O2 levels are not low, don’t come to the hospital,’” he said. “We don’t know how much harm that caused.”
He said options for getting more accurate readings for darker-skinned patients, such as taking painful arterial blood samples when lighter-skinned patients can use a simple clip-on device for the same measurement, are “really just trading one bias for another.” Valley, a critical care pulmonologist, called the inaccuracy of the devices a huge problem for clinical care. “I’ll be honest, sometimes I don’t know what to do in the hospital,” he said.
He advised patients with darker skin to question their pulse oximeter results and speak with their physicians, especially if they feel poorly or see any drop in oxygen levels.
“There is in my opinion only one fix,” he added. “We need pulse oximeters that work as well in Black patients as they do in white patients.” Many biomedical engineers have said that fixing the devices is not difficult technically; the issue is in getting the devices tested and approved, and having hospitals replace tens of thousands of the devices, which are more costly than the consumer versions. The FDA continues to monitor the issue and work with device manufacturers, and will provide updates to consumers as new information is available, an FDA spokesperson told STAT. Improved devices that use more wavelengths of light for better results in all skin tones are available, but not in wide use.
Valley said he was also concerned to see in the new study that while Asian patients had inaccuracies in their oxygen levels, they did not experience delays in treatment at the same rate as Black and Hispanic patients. “I think that’s as concerning as the pulse oximetry findings,” he said. “It suggests we are treating people differently.”
Fawzy, the study author, said that due to a small number of Asian patients in the study, there may not have been enough statistical power to detect whether Asian patients experienced delays. More and larger studies are needed on Asian populations; an earlier study of undetected low oxygen levels found, in contrast to the new study, that measurements of Asian patients more closely resembled those of white patients.
One issue, Valley said, is that race is an imperfect proxy for skin color. Better studies might compare patients by actual skin color and not racial groupings.
Utibe Essien, an assistant professor at the University of Pittsburgh School of Medicine who works on issues of equal access to treatment, said addressing poor oxygen measurement in some racial and ethnic populations was long overdue and that excuses should not be made because of the cost of acquiring and using more accurate devices. He said he was frustrated by the lack of investment and urgency to address a problem that affects such a large percentage of the population and may be a matter of life and death.
“This doesn’t just matter to provide equity,” Essien said. “It actually affects treatment.”
Biased pulse oximetry is similar to the race- and ethnicity-based cutoffs for lung and kidney function that led physicians to underestimate disease severity and limit access to transplants or other therapy for many patients in marginalized groups, wrote a group led by Valeria Valbuena, a general surgery resident at the University of Michigan, in an editorial accompanying the new study. Valbuena called for hospital systems and clinicians to address the “historical neglect of” and “diminished concern for” patients of color by insisting on purchasing medical devices that work well in all patients.
“The decision to do nothing about a faulty device is a human one,” the authors wrote, “and one that can and should be corrected.”