ABSTRACT Our healthcare system does not treat patients equally: some receive poorer quality care based on their racial/ ethnic identity, social class, health literacy, age, gender or medical conditions such as obesity, sickle cell disease, and substance use disorders. While there may be many factors that contribute to these disparities, implicit bias among clinicians is one important factor that remains difficult to measure. Clinicians may convey and acquire bias towards patients from each other when communicating orally or when writing/reading medical records. Although medical records are an integral method of communicating about patients, few studies have evaluated patient records as a means of transmitting bias from one clinician to another. Our own preliminary work has found that language used in medical records has a direct influence on subsequent clinicians who read the notes, both in terms of their attitudes towards the patient and their medication prescribing behavior; that there are racial disparities in the use of stigmatizing language; and that this language reflects actual attitudes held by the clinicians. This is an important and overlooked pathway by which bias can be passed from one clinician to another, further impairing healthcare quality for the individual patient as well as exacerbating disparities overall for those who are stigmatized. Our overarching goal is to improve the quality and equity of care for persons who are disrespected and stigmatized in healthcare. The specific aims of this proposal are: (1) to define and develop a taxonomy of stigmatizing language in medical records using qualitative analyses with input from clinicians and patients across 5 different medical specialty areas, (2) to measure and test for bias in the use of stigmatizing language in medical records using natural language processing (NLP) that can efficiently detect and quantify stigmatizing language in electronic patient records, and then (3) to reduce stigmatizing language in medical records using proven implementation science (IS) strategies (stakeholder engagement, education, audit and feedback, and champions) to motivate and enable health system change. Every encounter with a patient is documented in the medical record. Because language may have a powerful role in influencing subsequent clinician attitudes and behavior, attention to the language used in medical records could have a large impact on the promotion of respect and reduction of disparities for stigmatized groups.

PROJECT NARRATIVE Medical records are an integral method of communicating about patients and yet the language clinicians use to describe patients may reflect their own biases, be stigmatizing and adversely impact subsequent care. The specific aims of this proposal are to develop a taxonomy of stigmatizing language in medical records, to measure and test for bias in the use of stigmatizing language in medical records using natural language processing (NLP), and then to reduce stigmatizing language in medical records using proven implementation science (IS) strategies. Because language may have a powerful role in influencing subsequent clinician attitudes and behavior, this project could have a large impact on the promotion of respect and reduction of disparities for stigmatized groups.