Negative Patient Descriptors: Documenting Racial Bias In The Electronic Health Record

Area of Science:

• Medical Informatics
• Health Disparities Research
• Computational Linguistics in Healthcare

Background:

• The communication of racism and bias within medical records remains under-researched.
• Electronic Health Records (EHRs) are a critical source of patient information, yet may contain biased language.
• Understanding how patient race/ethnicity influences clinical documentation is crucial for addressing health inequities.

Purpose of the Study:

• To investigate whether healthcare providers' use of negative patient descriptors in medical records varies by patient race or ethnicity.
• To identify potential biases in clinical documentation within an urban academic medical center.
• To explore the association between patient race/ethnicity and the presence of stigmatizing language in EHRs.

Main Methods:

• Analysis of 40,113 history and physical notes from 18,459 patients (January 2019-October 2020).
• Utilized machine learning techniques to identify sentences containing negative patient descriptors (e.g., resistant, noncompliant).
• Employed mixed-effects logistic regression to assess the odds of negative descriptors based on patient race/ethnicity, controlling for covariates.

Main Results:

• Black patients had 2.54 times higher odds of having at least one negative descriptor in their history and physical notes compared to White patients.
• The study identified a statistically significant association between patient race and the use of negative descriptors in EHRs.
• Findings indicate a pattern of potentially biased language disproportionately affecting Black patients.

Conclusions:

• The presence of stigmatizing language in EHRs is a significant concern.
• Biased language in medical records may contribute to the exacerbation of racial and ethnic healthcare disparities.
• Further research and interventions are needed to mitigate bias in clinical documentation and promote health equity.