In 2020, a study published in the Proceedings of the National Academy of Sciences (PNAS) made waves in reporting that black newborns were more likely to survive when under the care of black physicians. In a report published by Do No Harm, we speculated that the correlation was spurious. Analysis from economists George Borjas and Robert VerBruggen confirmed our suspicions, and author correspondence retrieved through public-records request suggests that the study conclusion may not have been an oversight but an effort to preserve preferred narratives.
In penning a defense of DEI in medicine, the editors at Newsweek acknowledge that the PNAS study has been debunked. But they cite three other studies that claim to find benefits of racial diversity in medicine, arguing that, based on those studies, there appears to be promise for the hypothesis that racially concordant care is beneficial for minority patients.
That conclusion is incorrect.
First, those three studies must be properly situated within the broader literature. There have been — to our knowledge — six systematic reviews of race concordance in medicine. Systematic reviews consider what the weight and quality of evidence indicate about a research question. When it comes to race concordance, five of six reviews conclude that evidence does not support the concordance hypothesis, while the sixth features easily identifiable methodological errors.
Further, the three studies cited in defense of DEI warrant deep skepticism. As The Economist observed in reporting on the debunking of the PNAS study, “science that fits the zeitgeist sometimes does not fit the data.” We have previously reviewed each of these studies. The problems with all three are fairly obvious once you consider all of the facts.
The first study mentioned involves an experiment in Oakland where black patients were randomly assigned to one of six black or eight non-black physicians before agreeing to undergo preventative health screening tests.
In the first stage of the experiment, patients are shown a photo of a doctor on a tablet and asked about whether they would like to participate in four preventative health screenings. Before meeting with the doctor, patients state which of the screenings they would like to receive. At this stage of the experiment — which provides the strongest, least corruptible test of the hypothesis — willingness to participate in the preventative screenings does not vary according to the race of the doctor on the tablet.
Differences in the Oakland experiment are only observed in the second stage of the experiment, when the doctors depicted on the tablets visit the patients and attempt to persuade them to receive the screenings. At that stage, more patients assigned to visit with a black doctor and who had initially declined the screenings relented and accepted three of the four screenings compared to patients assigned to non-black doctors. The problem is that the second stage of experiment isn’t a test of race: It’s a test of the power of persuasion among six physicians who happen to be black compared to eight who are not. Whereas race would have been salient in the first stage (when patients only saw a photo), it’s impossible to decode whether it was a factor at all in the second stage. Moreover, the study acknowledges that “[t]he doctors, subjects, and field staff were not informed that doctor race was being randomized, though they could have inferred it” (p. 4082). As we wrote in our critique, “If the doctors were aware that the study was examining the effects of racial concordance, which seems likely, it is possible that sympathy with the racial concordance hypothesis would exceptionally motivate the 6 black doctors while undermining the motivation of the 8 non-black doctors to get patients to agree to interventions.”
The second study mentioned by Newsweek involves observing patients across the Military Health System and observing that, for black patients, transferring to a facility with a higher share of black physicians is associated with filling and renewing prescriptions with greater fidelity, as well as lower mortality.
Notably, the concordance effect (i.e., black patients being treated by black physicians) is never assessed directly. Data buried in an appendix clarifies the likely reason. That is, the effect the authors observe is driven by better outcomes among black patients who are served by non-black physicians in facilities with a higher proportion of black doctors. Suffice it to say, the idea that black doctors are no more effective than others at treating black patients but that their presence makes others more effective at treating black patients strains credibility. Precisely why they observe the pattern that they do is unclear, but making logical sense of the data indicates that a higher share of black doctors is the result of something that causes better relative outcomes for black patients rather than being the cause of those better outcomes.
The final study is the easiest to debunk. The authors analyze surveys taken by patients about their experience at a recent outpatient visit. The authors observe and dwell upon the fact that black patients are more likely to give the highest possible score to black doctors on a question about “the likelihood of your recommending this care provider to others” compared to non-black doctors. Notably, however, the average score given by black patients to black doctors doesn’t differ from the average score given to non-black doctors. In other words, the outcome is sensitive to considering only highest possible score versus all other possible scores and falls apart when considering the full range of outcomes. Other seemingly notable findings that are glossed over in the paper are that this same phenomenon is only observed among black patient-provider dyads and that there is no difference for white, Hispanic, or Asian patients when it comes to whether they would offer the most enthusiastic recommendation of a racially concordant provider compared to other providers. Further, the survey features 10 questions, including ones seemingly more important than the recommendation question. For example, it asks about “information the care provider gave you about medications (if any)” and “instructions the provider gave you for follow up care (if any).” The recommendation question was the only one of the 10 in which black patients were more likely to give the highest possible score to black doctors compared to other doctors. Altogether, then, the study fixates on a conclusion that is sensitive to how survey outcomes are operationalized, glosses over the fact that the phenomenon is observed in only one group, and minimizes the fact that concordance benefit is observed in only one of 10 survey questions.
Overall, the evidence is clear: Merit ought to stand alone in determining access to medical school. But you don’t need to take our word for it — just ask black patients themselves. In multiple national surveys, the vast majority of black patients say they have no preference for the race of their doctor. Of course they don’t. Like anyone else, black patients are focused not on race but on getting excellent medical care. And, like anyone else at the bedside, they prioritize technical competence, thoroughness, clear communication, and overall quality far higher than demographics.
