Language in EHR Clinical Documentation May Perpetuate Implicit Bias

Providers should be careful to avoid judgmental language during EHR clinical documentation lest they reflect and perpetuate implicit bias, according to a study published in JAMA Network Open.

These findings come as ONC Information Blocking rule requirements call on providers to offer patient access to open clinical notes. The language in those notes—and all EHR documentation—matter, the data indicates, and clinicians need to be aware of this to ensure a good patient experience.

Researchers analyzed 600 encounter notes written by 138 clinicians and found major linguistic themes that represent negative and positive physician attitudes.

Most negative language was not explicit and fell into one or more of the following five categories: questioning patient credibility, disapproving of patient reasoning or self-care, stereotyping by race or social class, depicting the patient as difficult to work with, and highlighting physician authority over the patient.

Positive language was more likely to be explicit and fell into these six categories: direct compliments, expressions of approval, self-disclosure of physicians’ positive feelings about the patient, blame minimization, personalization, and emphasizing patient authority for their own health decisions.

When physicians document personal sentiments regarding a patient within the EHR, they may influence the attitude and behavior of clinicians—or patients—who read the notes.

For instance, the study authors noted that a poor patient-provider encounter may lead a patient to believe that they are not receiving high quality care, putting them at risk for distrusting or disengaging from care.  

If the provider uses stigmatizing language within the EHR to characterize those patients, each subsequent clinician may treat them according to the impressions expressed by the previous clinician, furthering compounding a patient’s negative perception of their care quality.

“Negative feelings may stay with the patient when moving between clinicians, eliciting their past negative emotions and experiences and transferring it to other clinicians, creating self-fulfilling prophecies and confirming stereotypes,” the study authors explained.

“The consequences of this self-fulfilling prophecy may be documented repeatedly in the medical record, perpetuating bias and inequitable care,” they added.

The authors acknowledged that some of the statements they coded as conveying negative attitudes could be seen as relevant for future care purposes. However, clinical notes with negative depictions of patients may penalize them for a bad day and perpetuate implicit bias.

“These negative characterizations may come more from the clinician’s frustration or bias than from any inappropriate behavior on the patient’s part, compounding the injustice of clinicians, who hold testimonial power, using such language to describe people in their permanent records,” the study authors wrote.

Additionally, the study authors noted that the presence of compliments in patients’ records may influence clinicians to provide that patient with higher quality care compared to others.

What’s more, compliments of people from minority races may reflect underlying racism. For example, complimenting a Black patient for being “articulate” could, in fact, just be showing the provider’s low expectations and racism toward a specific racial group.

The study authors called for future research on how the use of stigmatizing language in the EHR varies by patient or clinician characteristics, and how this language impacts patient outcomes.

For instance, they noted that the linguistic patterns could be coded into natural language processing algorithms to allow large-scale quantification and categorization of potentially stigmatizing language within the EHR. This would allow researchers to study the implications of such language on patient care.

In addition, health systems could evaluate the prevalence of stigmatizing language to improve clinical documentation.

“Language has a powerful role in influencing subsequent clinician attitudes and behavior. Attention to this language could have a large influence on the promotion of respect and reduction of disparities for disadvantaged groups,” the study authors wrote.

Patients are increasingly accessing and reading notes in their own medical records due to 21st Century Cures Act requirements, furthering the need to eliminate stigmatizing language within open EHR notes.

Recent research published in the Journal of General Internal Medicine revealed that certain phrases can be off-putting to patients reading their clinical notes.

In many instances, it came down to word choice. Words such as “incorrect,” “obese,”
“wrong,” “anxious,” “depressed,” “inaccurate,” or “elderly” came up often as unfavorable to patients.

“I try and mirror the concept of what would it feel like if I was reading this out loud to the patient,” Leonor Fernandez, MD, a physician with Beth Israel Deaconess Medical Center, an assistant professor of medicine at Harvard Medical School, and an open notes researcher, told PatientEngagmenetHIT in a June interview.

“If I might wince and really substantially change the terms, because otherwise I feel they might be offended, then I shouldn't write that,” he explained. “We do this all the time when we talk to a patient in the room. We hopefully pick our words so that they're a little more meaningful, so that they're a little more affirming of the patient's strengths and what they're trying to do, or their fears.”