Doctors Know Stigmatizing Words Harm Patients, Yet Most Cannot Spot Them in Charts

Every day, clinicians at hospitals across the United States write thousands of notes describing their patients. Some of those notes contain words that quietly shape how the next clinician will treat the person on the other end of the stethoscope: descriptions of patients as manipulative, non-compliant, difficult, or not credible. A new multicenter survey published in the Journal of General Internal Medicine reveals a striking paradox at the heart of modern medical documentation. Nearly every physician and trainee surveyed believed that word choice matters and could bias care, yet almost none could reliably identify stigmatizing language when it appeared in front of them, and only a small fraction felt confident that their own notes were free of it.

The study, led by Julia Caton of Northwell Health and colleagues at Stanford School of Medicine, surveyed internal medicine residents, hospitalist faculty, and advanced practice providers at two large academic medical centers between April and September 2025. The researchers combined a traditional attitude survey with a knowledge test, asking participants to spot stigmatizing terms embedded in realistic snippets of clinical documentation. In total, 82 faculty and advanced practice providers responded, a 52 percent response rate, along with 63 residents, a 21 percent response rate. Because no validated instrument existed for this emerging field, the team built their survey from scratch, refining it through cognitive interviews designed to ensure that respondents at both institutions interpreted each question the same way.

The results expose a profound gap between intention and ability. Ninety-one percent of faculty and advanced practice providers and 92 percent of residents reported noticing stigmatizing language when reading clinical documentation. Roughly nine in ten said they considered how their own word choices might bias other clinicians toward or against a patient, and large majorities also considered the impact on patients and families who might read the notes. Yet when asked how confident they were that their documentation consistently avoided stigmatizing language, only 10 percent of faculty and advanced practice providers and 24 percent of residents reported being very or extremely confident. The clinicians knew the problem existed and believed it mattered, but they did not trust themselves to solve it.

The knowledge test made the scale of that distrust concrete. Participants were asked to identify 11 stigmatizing terms hidden across five documentation excerpts, with the terms selected through a literature review to capture a range of stigmatizing sentiments. Only 4 percent of faculty and advanced practice providers and 3 percent of residents correctly identified all 11. Mean scores were nearly identical between the two groups, with faculty and providers averaging 61 percent correct and residents averaging 59 percent, a difference that was not statistically significant. Experience, in other words, conferred no advantage. Senior attending physicians performed no better than first-year trainees at recognizing language that their own profession’s literature has repeatedly linked to worse care.

Recognition varied dramatically from term to term, and that variation may be the study’s most instructive finding. Terms such as non-compliant and claims were widely recognized as stigmatizing, while denies and the use of scare quotes around patient statements were far less frequently flagged. This pattern suggests that many clinicians have absorbed a short list of forbidden buzzwords without internalizing the underlying principles of patient-centered writing that would generalize to unfamiliar contexts. Education that simply teaches clinicians which words to avoid, the authors argue, will fail; effective training must convey the reasoning behind patient-centered language so that clinicians can navigate novel situations on their own.

The downstream consequences of stigmatizing documentation are not hypothetical. A 2018 study found that clinicians exposed to biased clinical vignettes developed more negative attitudes toward patients and became less willing to prescribe adequate pain management. More recently, researchers demonstrated that biased language used during verbal handoffs impaired clinical recall among medical students and residents and was associated with decreased expressions of empathy. Multiple studies have shown that stigmatizing language appears more frequently in the charts of Black patients, women, patients with substance use disorders, and patients with public insurance, raising the possibility that documentation practices actively perpetuate existing healthcare disparities. The medical record, once written, becomes a durable transmission channel for bias, read by every subsequent clinician who touches the case.

Open notes have raised the stakes considerably. Under the 21st Century Cures Act, patients in the United States can read their clinical notes immediately, and a recent study found that 10.5 percent of patients reported feeling offended or judged by something they read in a clinician’s note. A note is no longer a private communication between professionals; it is a document the patient may read within hours, one that can erode trust precisely when trust is most needed for treatment to succeed. The researchers note that this shift transforms documentation word choice from a stylistic quibble into a core clinical competency, one that the Accreditation Council for Graduate Medical Education implicitly recognizes through its Patient- and Family-Centered Communication milestone.

Faculty behavior adds another layer to the problem. Seventy-one percent of faculty said that avoiding stigmatizing language was very or extremely important for trainees, yet only 17 percent reported often or always providing feedback when they noticed stigmatizing language in resident notes. The most commonly cited barrier was lack of time, selected by 66 percent of faculty, followed by lack of prior training in giving feedback on biased language, cited by 29 percent, and lack of prioritization, cited by 26 percent. Documentation practices have traditionally been learned implicitly, through observation and the internalization of unspoken norms, rather than through explicit instruction. If supervisors rarely correct biased language due to time pressure and their own uncertainty, harmful habits pass unchallenged from one generation of clinicians to the next.

The study’s free-text responses revealed genuine tensions that any intervention must confront. Respondents worried about over-policing language and suppressing clinically meaningful information; a blanket prohibition on documenting that a patient refused a treatment, for example, could obscure safety-relevant events. Others pointed to conflicts with standardized terminology: terms such as obesity carry specific ICD-10 codes, and substituting alternative phrasing without guidance could affect reimbursement, risk adjustment, and disease capture in administrative datasets. Documentation simultaneously serves as a communication tool, a legal record, a billing instrument, and a quality measurement input, and optimizing language for one function can create friction in another. The authors argue that education must therefore frame patient-centered writing not as word prohibition but as a skill for achieving accuracy, respect, and clinical utility simultaneously, delivered with psychological safety and acknowledgment of legitimate gray zones such as direct quotations.

The findings arrive at a moment of technological inflection. Ambient listening technologies and large language model-assisted note generation are rapidly entering clinical workflows, and if these tools are trained on existing clinical notes, they risk encoding and perpetuating the very stigmatizing patterns that educators are trying to eliminate. The researchers have begun piloting microlearning modules for faculty and residents, developed an infographic for broader dissemination, and are pursuing electronic health record integration initiatives, including revised templates and automated flags that suggest alternatives without adding to faculty workload. Their data suggest the cultural groundwork is already laid: clinicians overwhelmingly agree that language matters. What remains is to build the skills, the feedback systems, and the technological safeguards that turn that agreement into notes that inform without harming, and that treat every patient reading their own record with the respect the profession claims to owe them.

Subject of Research: Clinician knowledge and practices regarding stigmatizing language in clinical documentation

Article Title: Hospital-Based Clinicians’ Knowledge and Practices Regarding Stigmatizing Language in Clinical Documentation: A Multicenter Survey

Article References: Caton, J., Sun, B., Steele, N., Santiago, C., Antara, F., Hom, J., & Dougherty, R. (2026). Hospital-Based Clinicians’ Knowledge and Practices Regarding Stigmatizing Language in Clinical Documentation: A Multicenter Survey. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10803-x

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10803-x

Keywords: stigmatizing language, clinical documentation, electronic health records, medical education, patient-centered language, healthcare disparities, open notes, internal medicine, medical bias, clinical notes, faculty development, survey research