Black patients were more than twice as likely to have their patient behavior and history characterized in negative terms compared to white patients, data from electronic health records (EHRs) revealed.
This higher likelihood of having a negative descriptor in their EHRs -- with terms such as "resistant," "noncompliant," or "agitated" -- showed up even after adjusting for sociodemographic and health-related characteristics (adjusted odds ratio [aOR] 2.54, 95% CI 1.99-3.24), reported Michael Sun, a third year medical student at the University of Chicago, and colleagues, writing in .
"We developed this machine-learning program to analyze the medical record, and look at what physicians and healthcare providers were actually saying about patients," Sun told ѻý.
While the study does not capture the potential impact of negative descriptors on patients' medical care, Sun's group highlighted that found explicitly stigmatizing language when written into hypothetical chart notes, "negatively affected respondents' attitudes toward the patient and resulted in less aggressive pain managment plans."
The authors also noted that because EHR notes are so regularly accessed by other care providers, there is a possibility of "recommunicating and amplifying potential biases."
Study Details
Researchers used a machine learning approach to identify potentially stigmatizing language in EHR data at a large urban academic medical center in Chicago (broken down by emergency department, inpatient, or outpatient setting). Patients coded for dementia were excluded from the study, as researchers noted that "negative descriptors may be applied to them more frequently because of the nature of their illness."
The study zeroed in on the history and physical note portion of the record as it provides the narrative about a patient, the authors said.
Sun's team analyzed 40,113 notes in the records of 18,459 adult patients taken from January 2019 to October 2020. Mean age of the patients was about 48, and 56% were women. About 61% of patients were Black, 30% were white, and 6% were Hispanic or Latino.
Overall, 8.2% of patients had at least one negative descriptor in the history and physical notes portion of their EHR. Notes in the outpatient setting had a lower chance of using negative descriptors (aOR 0.37, 95% CI 0.31-0.45) than inpatient notes, the authors said.
Both Medicaid (aOR 2.66, 95% CI 2.08-3.40) and Medicare enrollees (aOR 2.09, 95% CI 1.57-2.75) were more likely to garner a negative descriptor compared with patients covered by private or employer-based insurance.
Interestingly, unmarried patients had more negative descriptors versus married patients (aOR 2.12, 95% CI 1.70-2.65).
The study also found that increased score was tied to higher odds of a negative descriptor (aOR 1.11, 95% CI 1.07-1.15).
This may suggest, said Sun, that "patients who are more medically complex may be more frustrating to treat or ... patients with public insurance might have poorer access to healthcare, have poorer access to the medication," and may be less likely to follow-up on needed healthcare compared with someone who has more resources.
The study's limitations include limited generalizability and "limited racial and ethnic heterogeneity," which "prevented further disaggregation by race or ethnicity" so additional races were not included in their analysis.
The authors also noted that about 17% of the sample "may have been prone to selection bias."
Room for Change
Sun added that even if it takes a few more words, practicing "compassionate documentation" has practical uses.
"To say, 'This patient has poor medical literacy, or is not English-speaking or reading' ... that is so much more useful for future healthcare providers to then adapt to and respond to, than it is to say 'noncompliant' or 'non-adherent,'" he said.
The authors also found that notes written after March 1, 2020 (aOR 0.82, 95% CI 0.70-0.96) had lower odds of a negative descriptor than those written earlier in the study period, a finding that Sun said he found one of the most interesting.
With the heightened stress of the pandemic one might predict an increase in the use of negative descriptors after March 2020, Sun said.
However, it's possible that because of the activism during the summer of 2020 following the murder of George Floyd and the heightened awareness of the ways the pandemic increased current health disparities, "healthcare workers were more sensitized and paying more attention," he added.
"I think that points to an ability and capacity to change moving forward," Sun said. "It's not some sort of hopeless endeavor -- changing language. And it's not as hard as people think."
Disclosures
Sun disclosed support from the University of Chicago Medicine's Center for Healthcare, Delivery Science, and Innovation.
Other co-authors disclosed support from the National Heart, Lung, and Blood Institute, the Chicago Center for Diabetes Translation Research (funded by the National Institute of Diabetes and Digestive and Kidney Diseases).
Primary Source
Health Affairs
Sun M, et al "Negative patient descriptors: documenting racial bias in the electronic health record" Health Aff 2022; DOI: 10.1377/hlthaff.2021.01423.