‘Mail-Order Bride’ or ‘Very Obese’: Bias in Resident Patient Handoffs

Stigmatizing language — words that perpetuated stereotypes or suggested blame or skepticism — was found in 23% of audio recordings of patient handoffs by residents on inpatient general medicine and general pediatrics teams, a cross-sectional study found.

In 302 verbal patient handoffs, negative stereotypes were perpetuated at least once in 11% of nightly handoffs, according to Austin Wesevich, MD, MPH, of the University of Chicago, and co-authors. Blaming patients for their symptoms was identified at least once in 13% of handoffs, and casting doubt on patient reports and experiences was identified at least once in 5% of handoffs.

Overall, 23% of handoffs (70 of 302) had occurrence of any of the three kinds of biased language, Wesevich and colleagues wrote in a JAMA Pediatrics research letter.

Black race was the only demographic factor associated with bias overall (OR 1.7, 95% CI 1.0-3.0, P=0.049) and with negative stereotypes specifically (OR 3.5, 95% CI 1.6-7.8, P=0.002). BMI equal to or over 30 or being in the 95th percentile of weight was the only significant factor associated with blame specifically (OR 2.2, 95% CI 1.1-4.5, P=0.02).

“There’s all sorts of phrases that are passed along, that are normalized, and I think that’s what we’re trying to … expose and look at as objectively as we can, is ‘what are the ways in which people speak about patients?'” Wesevich told MedPage Today. “And it’s not that, in general, you have unfiltered language; you’re using it in ways that are disparate across patient populations. And that’s where it really points to a problem.”

Perpetuating negative stereotypes was defined as residents repeating negative stories or descriptors heard about the patient or the family. Examples included referring to an immigrant patient as a “mail-order bride” or another patient as “verbally and physically aggressive.”

Examples of blaming language included attributing a patient’s car accident and subsequent spinal fracture to being “drunk” or leakage from a patient’s ileostomy site as due to being “very obese.” Examples of doubting language including doubt that a patient’s self-described 10-out-of-10 pain could be due to bunions or describing another patient as having “a lot of malingering-type behavior.”

“The fact that we see any biased remarks, but especially significant association between bias and Black patient race, demonstrates the persistence of bias within the medical field,” commented Michael Sun, MD, also of the University of Chicago, in an email to MedPage Today.

“While a perhaps less generalizable study design, it may be a useful way for institutions to evaluate and identify bias among themselves to provide targeted feedback,” added Sun, who led a 2022 study on negative patient descriptors in electronic health records but was not involved in Wesevich’s study.

While a number of studies have evaluated written electronic health records for biased or stigmatizing language, fewer have examined verbal communication. “People should be even less filtered, or likely less filtered, with spoken communication,” Wesevich said. “And I think that’s something that was really interesting about this study: all of these people knew they were being recorded.”

“There’s lots of signals that the use of biased phrases is not happenstance, or equally distributed across patients,” Wesevich said. He also noted that even if a certain type of language or descriptor may not seem biased on the surface for one individual patient, “the fact that they were used at a higher prevalence for certain subpopulations makes you wonder, is it really just an objective description of how a patient is doing, or is it the clinician’s perception of that patient coming across?”

His group transcribed and coded 52 audio recordings of 302 individual patient handoffs by residents in general medicine or pediatrics departments based on the incidence of pre-determined types of bias: perpetuating negative stereotypes, blaming the patient for their symptoms, and casting doubt on patient reports and experiences (patients’ or their family’s reports).

Of the general medicine patients, 48% were Black persons, 52% were female, 39% had a BMI equal to or over 30, and 69% were publicly insured. Of the pediatric patients, 43% were Black children, 51% were female, and 71% were publicly insured. Patient factors evaluated for links to bias included Black race, Latinx ethnicity, female sex, obesity, insurance status, non-English speaking, psychiatric disorder, general medicine team, alcohol use, and drug use.

Prior to reviewing recordings, Wesevich said researchers used a dictionary of examples of stereotyping, blaming, or doubting language based on previous literature about bias. Recordings were double-coded by two attending physicians, with a third as a tie-breaker as necessary, he said. Demographic variables were determined via a Duke database and a review of medical records. Residents were aware of the recordings taking place and had been told ahead of time in an email that the study was analyzing patient handoff quality.

Limitations included the observational, single-center study design, potential for a Hawthorne effect (in which people modify their behavior because they know they are being observed), and grouping of racial categories for the multivariable logistic regression models. They added that the study was powered enough only to look at differences for Black, Latinx, or white patients.

Future studies, they wrote, are needed to identify the effects of such biased language on handoff recipient clinicians and the quality of care delivered subsequently.

Please enable JavaScript to view the

comments powered by Disqus.