Access and Equity
What Western Health Content Gets Wrong About African Patients
By Miracle Olajuyigbe 5 min read
Sickle cell trait is carried by roughly one in twelve African Americans. In Nigeria, it is closer to one in four.
Almost every piece of patient education written in English assumes the first number is the only one that exists. And here is the part that should bother you more: most of it does not mention either.
That omission is not a diversity oversight. It is a clinical error with a measurable consequence, and it runs through far more health content than anyone producing that content realises.
The case that makes this concrete
HbA1c is the standard test for diagnosing and monitoring diabetes. It works by measuring how much sugar has stuck to your haemoglobin, and it assumes your red blood cells live a normal lifespan of about 120 days.
Sickle cell trait shortens that lifespan. Less time in circulation means less time collecting sugar, which means the test reads low.
In 2017, a study in JAMA analysed data from two large community cohorts of African American adults, the CARDIA study and the Jackson Heart Study, nearly 8,000 people in total. Participants with sickle cell trait had lower HbA1c at any given glucose level than participants without it.
The important finding is not the size of the gap. It is what the gap does to the test’s usefulness. The researchers measured how well HbA1c identified people who met glucose-based criteria for prediabetes or diabetes. Without the trait, it performed reasonably. With it, performance dropped, and dropped sharply when measured against a two-hour glucose tolerance test.
In plain terms: for people carrying sickle cell trait, the standard diabetes test is meaningfully worse at finding diabetes.
Now hold that beside the prevalence figures. In the United States this affects a minority within a minority. In Nigeria it affects roughly a quarter of the adult population. Ghana, Cameroon, the DRC and Gabon are in similar territory.
So a Nigerian patient reads an article about HbA1c, written for an American audience, that tells them below 6.5 means they are fine. There is a one in four chance the test is systematically underselling their glucose. Nothing on the page mentions it. They go home reassured, and the clock keeps running.
It is not just the content. It is the instruments.
Content inherits its assumptions from the tools it describes, and the tools have the same problem.
In December 2020, a research letter in the New England Journal of Medicine compared pulse oximeter readings against direct arterial blood measurements across two large groups of hospitalised patients. Among patients whose oximeter showed a comfortable 92 to 96 percent, the researchers looked for cases where the true arterial oxygen was actually below 88 percent, which is genuinely dangerous.
In the University of Michigan group, that hidden low oxygen turned up in about 12 percent of readings from Black patients and about 4 percent from white patients. In the larger multi-hospital group, roughly 17 percent against 6 percent. Around three times the rate, in a device that clips onto a finger and is used everywhere, including in home monitoring programmes.
The oximeter works by shining light through tissue. Skin pigment affects how that light behaves. This was documented decades ago and largely ignored, because the calibration populations did not surface it.
So when you write “check your oxygen levels at home and call your doctor if it drops below 92,” you have written a sentence that is more reliable for some readers than others, and the page says nothing about it.
The pattern underneath
Once you notice it, you see the same shape repeatedly.
A reference range, a device calibration, or a screening threshold is established in one population. It becomes the default. It gets written into guidelines, then into software, then into patient content. By the time it reaches a blog post, the population it came from has vanished from view entirely, and what is left is a number presented as universal.
Nobody chose this. It is an accumulation of small defaults, each reasonable at the time, compounding into content that quietly works better for some readers than others.
What this means if you produce health content
I am not arguing for a paragraph about health equity at the bottom of every article. That is a gesture, and gestures are cheap.
Four things that actually help.
Name the population behind any number you publish. If a threshold, a normal range, or a screening cut-off came from a particular study group, say so when it matters. “Below 5.7 percent is normal” is a different sentence from “below 5.7 percent is the standard threshold, though the test can read low in people carrying sickle cell trait or certain other haemoglobin variants.” The second one costs you a clause and saves somebody a missed diagnosis.
Check whether your advice assumes a health system. “Ask your doctor to order a fructosamine test instead” is useful in London and close to meaningless where the test is not available and the appointment takes four months. You do not have to solve that. You do have to notice when you have written a sentence that only works in one kind of country, and offer something that works in both.
Ask who your content actually reaches, not who it targets. English-language health content is read globally regardless of who commissioned it. A US-focused diabetes article will be found by readers in Lagos, Nairobi, Karachi and Manila. The traffic data will tell you this if you look. Most teams never look.
Get someone to review who has practised somewhere else. This is the part you cannot research your way around. The failure modes I have described are obvious to a clinician trained where sickle cell trait is common and invisible to one trained where it is not. Neither is a better doctor. They have simply seen different things.
Why I keep coming back to this
Partly because it is the clearest example I know of accurate content still failing the reader. Every sentence in that HbA1c article is true. It is true and it does not apply, and the reader has no way to tell the difference.
And partly because it is fixable at almost no cost. One clause. One extra question in the brief. A reviewer who has worked in a different system. The gap between content that serves most readers and content that serves nearly all of them is much smaller than people assume.
Mostly it stays open because nobody in the production chain was in a position to notice it was there.