EHR Clinician Notes Offer Glimpse into Implicit Bias in Medicine

A look at EHR clinician notes gives insight into implicit bias in medicine, with EHR documentation indicating clinicians have disbelief in Black patients over White patients, according to a study published in the Journal of General Internal Medicine.

Researchers at Johns Hopkins initially observed different EHR documentation terminology within EHRs for patients with sickle cell disease that indicated the clinician did not believe certain reported patient pain levels. As a result, the researchers examined additional EHRs to observe if clinicians used similar language with other patients.

"We set out to see if we could identify linguistic mechanisms through which physicians communicate disbelief of patients in medical records and, if so, to explore racial and gender differences in the use of such language," Mary Catherine Beach, a faculty member at Johns Hopkins’ schools of Medicine and Public Health and a core faculty member of the Berman Institute of Bioethics, said in a statement.

Beach and her colleague, Somnath Saha, worked with a computer scientist and a linguist to identify three distinct language aspects in EHR notes that signaled physician disbelief of patients:

• Quotation marks around patient communication, such as a “reaction” to a medication
• Specific disbelief judgment words, such as “claims” or “insists”
• Evidentials, which is when the clinician describes a patient experience or symptom as hearsay

In that assessment of more than 9,000 EHR notes from over 150 physicians, Beach and Saha found that Black patients had a higher chance of containing at least one quote, at least one judgment word, and at least one evidential example than White patients.

"Our analysis of medical record language suggests Black patients are less likely to be believed by physicians,” Beach explained. “The bias reflected in those medical records may in turn affect care from future clinicians."

Gender-based health disparities did not occur among evidentials or judgment words, but female EHRs had a higher chance of containing at least one quote, said the researchers.

"We evaluated the prevalence of these features in over 9,000 notes in one clinic, then tested differences by race and gender,” explained Saha, MD, Johns Hopkins’ Berman Institute of Bioethics. “We found all three of these forms of language more often in the records of Black patients than white patients. Women's records were somewhat more likely than men's to have quotes, but not judgment words or evidentials."

"Some of this language reflects how clinicians are taught to document things, and there are reasons to use quotes and evidentials that don't necessarily cast doubt on what patients are saying. But if it's just benign word use, why would we see a difference in their application by patients' race and gender? That's what makes such language so insidious," Saha continued.

Because patient EHRs follow each patient to disparate healthcare facilities, these three language aspects could negatively impact patient care in the future and result in health disparities, Beach and Saha said.

"Clinicians know that patients are sometimes mistaken or even deceptive," said Beach. "But if we also know there is racial bias in the way patients' credibility gets assessed, we must revisit the certainty we have in our own impressions. We have to question ourselves before we question the statements of others."

Looking forward, Johns Hopkins Medicine asked Beach to speak to residents and current medical students at the university to address biased language impact on patient care and the value of patient trust.

“This is another potential mechanism for racial disparities in healthcare quality that should be further investigated and addressed,” the study authors concluded.