Covid-19 put medical research in a time warp. Sequencing the pathogen’s genome took a matter of hours. Tens of thousands of papers on treating and understanding it have sped from word processors to mass audiences within days. Highly effective vaccines were developed, approved and administered in months. And, for all of that, the virus may have taken countless extra lives on account of an equipment issue known and left unaddressed since at least 1990. Inevitably, this is about bias. According to age-adjusted APM figures, as of 2 March 2021, black Americans were twice as likely to have died from Covid-19 as white or Asian Americans. For indigenous Americans, the multiplier rises to 3.3; for Pacific islanders, 2.6; and for Latin Americans, 2.4. The reasons for these discrepancies are too numerous and wide-ranging to fully grasp, but as Covid effectively starves its victims of oxygen, it’s notable that 31 years earlier, a piece in Chest noted that the pulse oximetry target used for white patients on ventilators (92%) often resulted in hypoxaemia for black patients. Another study in 2005 found that pulse oximeters tended to overestimate blood oxygen saturation levels by several points in non-white people – by up to 8% at low ranges and from 1–4% around the vital clinical intervention points of 88%, 90% and 92%. Although even slight overestimations of oxygen saturation in this range could put patients at risk of severe organ damage or death, neither study appears to have impacted either device design or clinical practice.
Certainly, Michael Sjoding, an assistant professor of pulmonary and critical care and hospital medicine at the University of Michigan Medical School in Ann Arbor, had no idea that there was such a glaring flaw in one of his most important diagnostic tools. Then the first surge of Covid brought him an influx of breathless patients from nearby Detroit. As 78% of Detroit’s population identify as black or African American, compared with 7% in Ann Arbor, this radically altered the hospital’s patient demographic, revealing some uncomfortable truths about Sjoding’s practice in the process.
“Honestly, when we were caring for these patients, we saw this phenomenon happening,” recalls Sjoding. “The pulse oximeter was reading a normal value, and then, just by chance, we had a confirmatory arterial blood gas [ABG] test drawn, and it was way off. We had no idea what to make of it.”
Like most of his peers around the world, Sjoding had been operating under the assumption that pulse oximeters were as accurate as numerous clinical studies suggested. Prior to Covid this caused few, if any, issues, because the population of those clinical studies looked a lot more like the residents of Ann Arbor than Detroit. Suddenly forced to apply expertise honed for a quite homogenous patient group to a much more diverse one, however, Sjoding and his colleagues became increasingly aware of a mismatch between their expectations and the reality on the wards. “I think because we were caring for so many more black patients than was typical, this was a thing we started noticing, which we weren’t really keen to,” he says.
Nonetheless, at first, Sjoding and his “puzzled” colleagues wondered whether this was a Covid-specific issue. If not for the unlikely intervention of medical anthropologist Amy Moran-Thomas and the literary magazine Boston Review, their interest in it may well have waned as their patient demographic normalised. In the course of her research, Moran- Thomas, a professor at MIT, unearthed the articles cited above, along with a curious tendency for medical practitioners and device designers to deny that their findings mattered in practice. As she notes in her August 2020 article ‘How a Popular Medical Device Encodes Racial Bias’, “Most oximeters on the market today were initially calibrated primarily for light skin, and they still often reproduce subtle errors for nonwhite people.”
Sjoding saw the piece. “She dug out this old research describing these potential disparities and discrepancies and postulated that it might be happening in the Covid pandemic,” he explains. “We were like: ‘This is happening in the Covid pandemic – we’ve seen it with our own eyes’.”
Investigating the problem
To prove it, Sjoding and his colleagues quickly launched a real-world study comparing pulse oximeter and ABG readings from patients hospitalised at the University of Michigan in 2020 as well as others across 189 hospitals in 2014–15. “We wanted to show that it’s still an important problem and probably remains very underappreciated,” he says.
The team found that pulse oximeters can provide misleading results in more than one in ten black patients. Specifically, black patients in the study had nearly three times (11.7%) the frequency of occult hypoxaemia (arterial oxygen saturation of less than 88% and a pulse oximetry (SpO2) reading of 92–96%) as white patients (3.6%).
In the context of a pandemic that has disproportionately affected ethnic minorities in Western countries, as well as the growing support for the Black Lives Matter movement, Sjoding’s study, published as a research letter in the New England Journal of Medicine, did what its predecessors couldn’t: it drew attention to the problem.
For Olamide Dada, who was working for the NHS Race and Health Observatory at the time, the findings were not exactly a surprise. “Just observing the impact that Covid-19 was having on ethnic minority groups, in my head it was, ‘Why haven’t people put two and two together and considered this as a potential contributing factor?’” she says.
Sjoding is similarly bemused that his work should have been so heralded. “Our publication, which has been quite well-received and has made a big impact, in some respects isn’t new,” he sighs. “The knowledge has been around for a while, so why wasn’t this corrected 20 years ago? And, moreover, here we are, practising physicians, and we’re not aware of this knowledge, or of the past studies demonstrating the bias could exist. It was not disseminated in any meaningful way.”
That looks like it might be changing. In March 2021, Dada, a newly qualified doctor and the founder of the UK charity Melanin Medics, authored the NHS Race and Health Observatory report “Pulse oximetry and racial bias: Recommendations for national healthcare, regulatory and research bodies”. A month earlier, the FDA issued a public warning about the devices’ limitations, although it did not explicitly mention racial disparities in their accuracy. Elsewhere, Sjoding has seen US universities update their medical curriculums to reflect issues with pulse oximeters in non-white patients, and numerous specialist societies and medical bodies have released statements and recommendations. The Anesthesia Patient Safety Foundation (APSF), for one, now advises that “clinicians should not make patient care decisions such as hospital or intensive care unit discharge on the basis of a single SpO2 value”.
But there are a lot of doctors out there. In the UK, at least, Dada believes much more needs to be done to reach them. Once again, she’s concerned that the wider profession is moving on from discussing the issue without properly addressing it. Although NHS England and the MHRA have published guidance informing the public of the problem, advising them to pay more attention to pulse oximetry trends over time than singular readings, there has been no equivalent update for doctors and nurses. “Yes, patients need to be made aware, but so do healthcare professionals,” says Dada. “And that’s one thing that I’d say actually hasn’t been amplified since the release of our report. If a patient is saying, ‘I am not confident in this device’s reading at this time,’ a doctor who isn’t aware of the fact that such significant deviations exist might just dismiss their concern. What’s the point of patients becoming more aware if doctors are not adapting or adjusting their practice accordingly?”
A more complete picture
The problem, as Dada sees it, is that doctors don’t understand how they can use Sjoding’s findings to better serve their patients. Currently, it’s easy to think of this as a problem without a solution. There’s no similarly cheap and non-invasive replacement for today’s devices, and abandoning them would be counterproductive. In a recent pre-print study, at-home use of pulse oximeters after a positive Covid test lowered the risk of death by 52% in South Africa. As the APSF also stresses: “It is potentially more harmful if the known bias in measurement related to skin tone resulted in a lack of confidence in pulse oximetry as a monitoring tool for patients with dark skin tones.” Depriving patients of diagnostics based on the colour of their skin is no way to address inequality.
So, doctors need to be sceptical of pulse oximetry readings without lacking confidence in pulse oximeters themselves. That’s not an easy balance. “One of the challenges is the practicality of what this means,” points out Dada. “What does it mean for our practice, for our patients, for us? In terms of how we’ve been practicing medicine, this is all we’ve known.”
For Sjoding, simply being mindful of those questions can be productive. “My sense is the pulse oximeter has become so ubiquitous, and we’ve become so comfortable using it, that we’ve forgotten that it’s not perfect,” he says. “Most of what we have in medicine is imperfect, but we’ve taken it for granted.” He’s not going to stop using tools that provide valuable information so rapidly. “But I have a much greater awareness that, when the pulse oximeter is reading a value like 92%, the range of what that could actually mean is much wider than I previously appreciated, particularly in black patients.”
Although Sjoding is now less reliant on the tool with which his specialism is arguably most associated, his new approach is still true to his training. “I think a lot of what we do in critical care is synthesise information,” he continues. “One or two pieces of data that you have may be imperfect, or most of the data, but when you synthesise it all together, you get a more complete picture, and you’re less likely to be influenced by one thing that’s off.”
Similarly, Dada notes that her threshold before doing an ABG for darker skinned patients with breathing difficulties is now lower than it might be otherwise. Of course, in the most strained scenarios, that might not count for much. Stretched resources might put anything other than pulse oximetry readings out of reach. In the early stages of the pandemic, pulse oximetry readings were prized as the ideal biomarkers for rapid clinical decision-making, meaning some patients were wrongly sent home or denied treatment that hospitals were equipped to give them.
That does not need to happen again. With investment and attention, pulse oximeters can be adapted and improved. As Moran-Thomas points out in a follow-up piece in Wired magazine, 1970s Hewlett- Packard ear oximeters were specifically designed to work the same way across all skin tones, body types and disease states. Similarly, Sjoding’s data does not indicate how different pulse oximeter brands and models perform, leaving open the possibility that certain designs in use today are more equitable.
Even before those questions are answered, or new models become available, however, Dada wants practitioners to realise that “there is something that can be done”. Like US and UK-based critical care trainees Daniel Colon Hidalgo, Olusegun Olusanya and Emily Harlan, who recently published a letter in The Lancet, they can press for change to how pulse oximeters are tested, approved and purchased. Like Dada herself, they can discuss the issue with their colleagues, using both research like Sjoding’s and individual case studies. Before that, “it can start with your interactions with your patients,” she says: “just sharing your awareness that these devices may not read as accurately for certain people and responding accordingly.” The few seconds that takes could change everything.
