Abstract:On November 29, 2020, Dr Susan Moore was admitted to the hospital after testing positive for COVID-19. Days later, in a video-recorded Facebook post from her hospital bed, she told the world that her physician had dismissed her shortness of breath and did not provide the care she needed. Dr Moore, who was well positioned as a Black woman and physician to understand the role of racism in health care, said in her post, "This is how Black patients get killed." She died 2 weeks later. 2 Dr Moore's experience reflects a particular form of racial prejudice in which Black patients are less likely than others to be believed or to have their concerns taken seriously. This bias in clinician assessments of patient credibility is a manifestation of testimonial injustice, a form of epistemic injustice that occurs when a speaker (in this case, a patient) receives an unfair deficit of credibility due to a prejudice on the part of the hearer (in this case, a clinician). 3 Testimonial injustice in health care has received less attention than other manifestations of racial bias, including stereotypes of Black patients as difficult, aggressive, or noncompliant, but has been well described by Black patients themselves. In focus groups about patient perceptions of respect in health care, we found that one of the most prominent experiences of disrespect for Black patients was being dismissed, disbelieved, or not taken seriously by their doctors. 4 To investigate whether these experiences were reflected in medical records, we used natural language processing (NLP) methods to identify credibility-related language in the electronic health record (EHR) and found that clinicians used language undermining credibility substantially more often among Black compared with White patients. 5 The study by Lee et al 6 largely confirms our findings and adds to the evidence of testimonial injustice in health care. To identify EHR language reflecting testimonial injustice, Lee et al 6 adopted a framework similar to ours, with 2 different linguistic concepts related to implied credibility: epistemic modality and evidentiality. 5 Epistemic modality relates to the degree of certainty a writer conveys about a piece of information. For instance, a clinician writing, "the patient claims to have 9/10 pain," implies hesitancy to believe that claim and thereby directly challenges the patient testimony. Evidential terms, on the other hand, may not directly imply doubt but rather attribute the information to its source (ie, the patient). This allows the writer to disclaim responsibility for whether that information is true. For instance, the statement "the patient reports having 9/10 pain" does not inherently cast doubt but does leave more room for the possibility that the testimony is false or exaggerated than would the declarative sentence "the patient has 9/10 pain." Readers may legitimately question whether evidential terms (words like says , reports , and describes ) always cast doubt on a patient's testimony. They do not, and we caution against the impulse to eliminate their use. Evidential terms serve important functions. First, they separate subjective (patient-reported) from objective (clinically observed) information, which is important for clinical synthesis and decision-making. For instance, it is important to note whether an irregular heart rhythm is reported by a patient or clinically documented. Second, the evidential terms endorses and denies are usually used to indicate the presence or absence of symptoms elicited through questioning rather than spontaneously offered by the patient, which may be important for distinguishing primary vs associated symptoms. Third, terms like believe may be more indicative of the patient's uncertainty about the information in question than the clinician's. For instance, the statement "Patient does not believe that hepatitis C is active" suggests that the patient may be uncertain and that the status of hepatitis should be verified. Changing that statement to "Patient does not have active hepatitis" communicates a finality that may lead to an active illness persisting undetected. With those caveats in mind, evidential terms certainly can reflect doubt, and the finding that they occur more commonly in the records of Black patients is indicative of racial bias in physician assessment of patient credibility. There is no other convincing rationale for racial differences in the use of these terms. In their analysis, Lee et al 6 thoughtfully addressed potential alternative explanations for their findings by controlling for note length and demonstrating racial equivalence in the use of language unrelated to credibility. Thus, while we do not believe that the use of evidential terms is inherently problematic, their differential use in the medical records -Abstract Truncated-

Testimonial Injustice: Linguistic Bias in the Medical Records of Black Patients and Women

Intersectionality and Testimonial Injustice in Medical Records

The Credibility Deficit Experienced by Black Patients

See Me and Believe Me: Causality and Intersectionality in Testimonial Injustice in Healthcare

Examining Linguistic Differences in Electronic Health Records for Diverse Patients With Diabetes: Natural Language Processing Analysis

Disparities in Documentation: Evidence of Race-Based Biases in the Electronic Medical Record

Race and Ethnicity and Clinician Linguistic Expressions of Doubt in Hospital Admission Notes

Race and Ethnicity and Clinician Linguistic Expressions of Doubt in Admission Notes

How to Reduce Stigma and Bias in Clinical Communication: a Narrative Review

Linguistic Stereotyping in Older Adults’ Perceptions of Health Care Aides

"Just Because You Have Ears Doesn't Mean You Can Hear"-Perception of Racial-Ethnic Discrimination During Childbirth

Black Women in Medical Education Publishing: Bibliometric and Testimonio Accounts Using Intersectionality Methodology

Measuring Implicit Bias in ICU Notes Using Word-Embedding Neural Network Models

Discriminatory and valuing communication behaviors in cardiology encounters

Physicians' Implicit Racial Bias and Verbal Dominance and Perceived Discrimination in Goals of Care Conversations (RP115)

The Lived Experiences of Racial Microaggressions for Black Individuals While Seeking Orthopedic-Related Care: A Qualitative Study

Preliminary evidence of linguistic bias in academic reviewing

Differences in Narrative Language in Evaluations of Medical Students by Gender and Under-represented Minority Status

Assessment of Gender-Based Linguistic Differences in Physician Trainee Evaluations of Medical Faculty Using Automated Text Mining

Analysis of Race, Sex, and Language Proficiency Disparities in Documented Medical Decisions