The political distortion of medical research has a sordid history, but it’s unfortunately not just a thing of the past. Today, a popular narrative has taken hold that a racist medical establishment is the reason that blacks have shorter life expectancies, worse clinical outcomes for many diseases, and even excess maternal and infant mortality. The claim is unsupported by evidence, however, and believing it won’t do anything to improve black patients’ health.
Search for the terms “racism” and “medicine” in the National Library of Medicine database, and thousands of scientific publications appear. Journalists and a growing number of doctors regard this as proof of medical discrimination. But most of these studies do not prove any causality; they merely document disparities in clinical outcomes and medical services for black Americans. Nonetheless, they increasingly serve to justify such discriminatory practices as preferentially reserving scarce Covid-19 therapies for blacks.
A rush to find racism typifies most of the many thousands of opinion pieces, original investigations, and review articles on the topic of clinical outcomes for black patients. That literature supports a media that has eagerly adopted the narrative of racism embedded in American health care. The result undermines the trust in medical care needed for successful patient- physician relationships and diverts scarce resources in combating a nonexistent factor in poor health outcomes.
The rules for conducting robust scientific research require scientists to try to disprove their own theories. One can never absolutely prove a hypothesis correct; one can only show that experiments fail to disprove it. The investigator should begin by doubting the hypothesis and do his best to disprove it with carefully designed experiments. Unfortunately, too many studies on medical racism are carried out by investigators who, following the prevailing political trend, set out to confirm their ideas of a racist health-care system. A biased experiment can easily lead to a desired outcome, and emphasizing some results while ignoring others can lead to a faulty conclusion.
Consider a 2022 research article in the highly regarded journal Health Affairs. Titled “Negative Patient Descriptors: Documenting Racial Bias in the Electronic Health Record,” the study uses automated systems to review more than 18,000 electronic medical records. It finds that “bad descriptors” are used in the medical records 2.54 times as frequently in the records of black patients compared with white patients. The headline: “Our findings raise concern about racial bias and possible transmission of stigma in the medical record.”
One might expect that the investigators found that physicians and nurses demeaned or disparaged patients based on their race. So what were the negative words that allegedly show bias? The most common was “refused.” If a patient refused treatment, refused recommendation for a procedure, refused allowing a blood draw, or refused a medication, the authors consider a physician’s or nurse’s writing about the refusal in the medical record an indicator of racism. But that word describes the patient’s actions—it is a fact, not an opinion. Other words that the AI robot found to indicate biases were “not adherent” and “agitated.” Health Affairs highlights this paper in its advertising, but its spuriousness can be discerned only with a close look into its methods and results sections.
Medical research is rife with similar studies. Another study, cited hundreds of times in the medical literature and published in the New England Journal of Medicine, used simulated patient cases to determine whether physicians would assess black patients with chest pain differently from the way they would white patients with similar complaints. The physicians treated black men exactly the same as white men but referred black women less often for cardiac procedures than white women (it is not clear why). Journalists and academics cite this study to support the idea of racism in cardiac care, but they rarely point out that men received equivalent care.
The widespread availability of large databases like the one used by the Health Affairs investigators allows studies of the difference between black and white patients’ utilization of procedures and health outcomes. A notorious example, published in the Proceedings of the National Academy of Sciences, used a database from Florida to show that black newborn babies had a greater chance of survival if they were treated by a black pediatrician. A dangerous conclusion might be that patients should seek out physicians matching their own race. If this study is correct, we could be on the path to medical apartheid.
But the study suffers from fatal mistakes. Any study using a large database to assess the cause of death should undertake a chart-level assessment of the circumstances of the patient’s death. A large database is often riddled with errors, as the various entries are made by administrative personnel who are rarely, if ever, trained in health care. It’s a game of telephone: the truth can become hopelessly muddled after multiple rounds. In this paper, the infant mortality data were never checked at the level of the patient’s chart, so it’s impossible to know which doctors actually cared for the patient during any acute event. Moreover, the authors had no way of determining the race of the physician of record besides scanning available photos, and almost 2,000 of the nearly 10,000 doctors in the sample had no photos (they were excluded from the analysis). Finally, it was not clear whether the infants who died had been referred from outlying hospitals to large medical centers because they were already critically ill. If so, the treating physician under whose care the patient died may have received a desperately ill baby with little chance of survival.
These flaws notwithstanding, the study was widely cited in the press as proof of the need for more black physicians. USA Today headlined its story: “Black babies are more likely to survive when cared for by Black doctors, study finds.” The coverage only fed the narrative that racism permeates American medicine.
One could go on dismantling studies like these. But the key finding lacking in any of the studies of racism in medicine is evidence that the countless diversity, equity, and inclusion trainings to which doctors are now subjected would alter patient outcomes. A vast gulf remains between methods of traditional medical research and these alleged remedies. In medical research, cures are proposed and then tested in two populations; if an improvement in outcomes results between the two groups, the therapy may work, and the “null hypothesis” is rejected. But with an antiracist approach to improving medical outcomes for black patients, a cure is proposed, consisting of expensive and time-consuming antiracist training, but the testing step is ignored. One simply must adopt the antiracist cure and implement it widely.
To the extent that medical research exaggerates racism in clinical outcomes, it does a disservice to identifying the real basis for discrepant results. Disparities aren’t always due to discrimination. If a genetic trait is the culprit—for example, increasing susceptibility to chronic kidney disease—then encouraging physicians to become activists will do nothing to improve patients’ outcomes. The failure of antiracist programs to do anything to improve clinical outcomes for black patients will only deepen the frustration of clinicians and the dismay of patients.
Doctors should conduct research and find treatments that work. They shouldn’t treat patients differently based on skin color. Doing so would undermine everything that physicians pledge when they first are called “doctor.”
Stanley Goldfarb is a professor of medicine and former associate dean of curriculum at the University of Pennsylvania’s Perelman School of Medicine and the author of the forthcoming book Take Two Aspirin and Call Me by My Pronouns.
https://www.city-journal.org/political-distortion-of-medical-research