Electronic Health Records Suggest Racial Bias in Assessment of Patients’ Claims

Black patients were more likely to have comments from their physician questioning their competence and sincerity in their electronic health records (EHR) when compared to White patients, according to a new study published in PLoS One. ¹

The study, led by Mary Catherine Beach, MD, MPH, aimed to address racial and ethnic disparities in health care. While researching treatments for a measure of pain management for people with sickle-cell disease—which disproportionately affects Black patients²—Beach witnessed, on numerous occasions, physicians questioning the severity of patient-reported pain.¹

Analysis of electronic health records suggest racial bias undermining patient claims in competency and sincerity. | Image Credit: adobe_thodonal.jpeg

“We noticed that when someone might write that the patient reported that they had 10 out of 10 pain, they would word it in such a way and include details that would undermine the credibility of that reported pain level,” Beach said in an interview with The American Journal of Managed Care^®.

“Stunned” by responses from Black community members indicating that their health concerns weren’t being taken seriously by physicians, Beach sought to investigate and discovered that physicians were documenting their disbelief in patient health records. Using natural language processing models (NLP), researchers were able to identify terms that started patient claims with words or phrases like “insists” or “is adamant about” their symptoms or is a “poor historian.”

“There is a way in which those words can also be used to imply that the person is just being difficult, like he's ‘adamant’ that he's going to leave the hospital today. Our NLP models were able to distinguish those two different ways of using the words ‘insists’ and ‘adamant,’” Beach explained. “So, we were only looking at the ones where the patient was either ‘adamant’ or ‘insisted’ that something was true.”

Racial Language Disparities in Patient Notes

The study analyzed 13,065,081 notes from EHRs between January 2016 and October 2023; this included the notes of 1,537,597 patients written by 12,027 clinicians from a large health system made up of 5 hospitals.

The majority of patients' notes assessed belonged to females (55.95%), and the mean age was 43.3 years. Moreover, 57.6% of patients were non-Hispanic White, 28.0% were non-Hispanic Black, 8.3% were Hispanic/Latino, and 6.1% were Asian. Amongst the physicians, 44.9% were attending physicians, 40.1% were physicians-in-training, and 15% were advanced practice providers. The analysis revealed that less than 1% of notes contained language undermining patient credibility (n = 106,523; 0.82%), and out of those, 62,480 (0.48%) contained language undermining patients’ sincerity, and 52,243 (0.40%) contained language undermining competence. While the percentage of notes questioning credibility may be small due to the sample size, the number of patients affected by those notes is still significant.

“I think that it is reassuring that not a high percentage of notes have this kind of language in them. In order for clinicians to actually write something that directly undermines someone's credibility, they have to really feel a little more comfortable going on record with that," Beach said. “But I think it's practically significant that even less than 1% have had this language, because this can happen to anyone, and it is very dangerous to not be believed and listened to and taken seriously when you're reporting your symptoms or your preferences."

Non-Hispanic Black patients still had higher overall odds of having notes written using terms that question their credibility when compared with non-Hispanic White patients (aOR 1.29, 95% CI 1.27-1.32).¹ The odds were similar for each term categorized and assessed for questioning credibility. On the other hand, there was no difference in undermining language between Hispanic/Latino patients when compared with White patients (aOR 0.88, 95% CI 0.83-0.93); while Asian patients had lower odds of language undermining sincerity (aOR 0.74, 95% CI 0.69-0.79), but no difference in competence when compared with White patients (aOR 0.95, 95% CI 0.89-1.01).

Furthermore, language that supported patient credibility, including terms like “good historian,” was also assessed and present in 33,706 (0.26%) notes. Black patients had lower odds of notes containing language that supported credibility when compared with White patients (aOR 0.82; 95% CI 0.79-0.85) whereas, Asian patients had higher odds than White patients of notes with language supporting their credibility (aOR 1.30, 95% CI 1.23-1.38).

Study Authors' Suggestions and Actionable Solutions

The study authors proposed that sometimes the use of “poor historian” by clinicians can be due to the clinician failing to collect a complete history because the patient could not “communicate it clearly,” in which case they suggest using different phrases like “patient unable to provide a complete history” or “patient is uncertain of some details,” thus conveying less blame or doubt by the physician.

“As stigmatizing language in medical records has been shown to negatively influence clinician attitudes and decision-making, these racial differences in documentation may negatively influence quality of care and outcomes and further exacerbate health inequities,” the study authors wrote. “This specific type of bias related to credibility assessment has been relatively overlooked and has therefore remained unchecked.”

The study limitations did not account for physician covariates (sex, age, and race/ethnicity), in addition to the limited number of terms of credibility phrases assessed by the NLP model. The study sample was limited to one hospital system from the mid-Atlantic region that treated a higher proportion of Black patients than the national average. Nurse notes were also not assessed due to their templated nature, but researchers encourage these areas for future study.

“Medical educators have taken up efforts to train future physicians to engage in more respectful, patient-centered documentation and to learn about the structural inequities and interpersonal attitudes that underlie biased language use in the EHR,” the study authors concluded. “We recommend that these curricula include content on testimonial injustice and suggest eliminating phrases such as ‘poor historian,’ as well as veiled insinuations about patient insincerity.”

References

1. Beach MC, Harrigian K, Chee B, et al. Racial bias in clinician assessment of patient credibility: evidence from electronic health records. PLoS One. Published online August 13, 2025. doi:10.1371/journal.pone.0328134

2. Data and statistics on sickle cell disease. Centers for Disease Control and Prevention. May 15, 2024. Accessed August 13, 2025. https://www.cdc.gov/sickle-cell/data/index.html