The equations that used race, and why medicine removed them

This topic separates applicants sharply, because the easy answer is available to everyone and worth almost nothing. Saying that race has no place in medicine is a slogan; explaining why a coefficient that made a regression fit better was nonetheless wrong is an argument, and it is the argument a committee is listening for.

Here is the version worth holding. The race variable was added because it improved prediction on average in the datasets available. It was removed because a variable that predicts on average can still misclassify a specific patient in a systematic direction, because the association it captured was a consequence of social conditions rather than a cause rooted in biology, and because the number it produced was used to gate access to treatment. All three of those have to be in a strong answer.

Kidney function is estimated rather than measured directly. The usual approach measures serum creatinine — a waste product of muscle metabolism — and puts it into an equation with age and sex to produce an estimated glomerular filtration rate. Because creatinine depends partly on muscle mass, anything that shifts muscle mass shifts the estimate.

When the widely used equations were developed, the study populations showed slightly higher average creatinine in Black participants, and a coefficient was added to correct for it. The effect was to raise the estimated eGFR for any patient recorded as Black — that is, to report better kidney function from the same blood test. The justification offered was average differences in muscle mass, which was never well substantiated and which, crucially, is not what race measures.

The consequence was clinical. eGFR thresholds gate referral to nephrology, eligibility for transplant waiting lists, and the dosing of many drugs. A higher reported eGFR could mean a later referral, a later place in the transplant queue, or a dose that was wrong. Analyses after the change found that removing the coefficient reclassified substantial numbers of Black patients into more severe stages of chronic kidney disease — patients who had, on the previous equation, been recorded as healthier than they were. In 2023 the transplant system went further and required programmes to recalculate Black candidates’ waiting time as though the race-free equation had always applied, restoring time many had lost.

Three claims do the work, and they are worth being able to state separately.

Race is a social category. It is not a proxy for genetic ancestry — there is more genetic variation within any racial group than between groups — and it is recorded inconsistently, often by observation or self-report, with no defined handling of mixed heritage. A variable that cannot be measured reliably has no business in an equation that decides transplant eligibility.

What the coefficient captured was consequence, not cause. Where average differences in a health measure track race, the usual explanation is the cumulative effect of unequal exposure, income, environment and care. Encoding that association into a clinical tool converts a symptom of inequity into a rule that perpetuates it, and does so with the authority of a laboratory number.

Average fit does not justify individual use. Even where a coefficient improves prediction across a population, applying it to an individual assigns them the average of a group they were sorted into by a social process. For a decision as consequential as transplant listing, that is the wrong unit of analysis.

The counterargument deserves a hearing: if a variable genuinely improves accuracy, removing it makes estimates worse for someone, and clinicians pointed out during the debate that the replacement equation is less precise for some patients. The task force’s answer was that a modest loss of average precision was worth eliminating a systematic, one-directional harm — and that the better fix is to measure what actually matters, which is why cystatin C, a marker less affected by muscle mass, is increasingly recommended alongside creatinine.

Spirometry. Lung function reference ranges have long applied race-specific adjustments, meaning a Black patient’s measured lung function was compared against a lower expected value — with consequences for diagnosis, disability assessment and occupational compensation. Professional bodies have moved toward race-neutral reference equations, and the transition is ongoing.

Obstetrics. A widely used calculator estimating the likelihood of a successful vaginal birth after caesarean included race and ethnicity terms that lowered the predicted success for Black and Hispanic patients, potentially steering them toward repeat caesarean. The developers released a revised version without those variables in 2021.

Pulse oximetry. This one is different in kind and worth flagging as such: there is no race coefficient to remove. The device itself overestimates oxygen saturation in patients with darker skin pigmentation, which came to prominence during the COVID-19 pandemic when it meant occult hypoxaemia going undetected and patients being missed for treatment thresholds. The fix is engineering and regulation, not deleting a variable — and knowing that difference is what shows a panel you understand the category rather than the slogan.

This arrives in three shapes. The direct one: "What do you know about race correction in medicine?" The values one: "How would you address health disparities as a physician?" And the disguised one — "Can an algorithm be biased?" or a question about artificial intelligence in clinical decisions.

For the direct question, use the kidney example and walk the three claims. For the values question, use this as the concrete instance rather than speaking in generalities. For the algorithm question, note that a model trained on unequal care learns to reproduce it.

Start with the National Kidney Foundation and American Society of Nephrology task force’s final report, which sets out both the reasoning and the dissent in the profession’s own words. The FDA’s communications on pulse oximeter accuracy cover the device problem, and the American Thoracic Society’s statement covers spirometry.

Two pieces here sit beside this one. How American healthcare is actually structured explains the system in which these tools gate access, and rural hospital closures is the geographic form of the same inequity. For the interview formats, see our US interview guides.