How to Reduce Stigma and Bias in Clinical Communication: a Narrative Review

Abstract

A growing body of literature demonstrates that healthcare providers use stigmatizing language when speaking and writing about patients. In April 2021, the 21st Century Cures Act compelled clinicians to make medical records open to patients. We believe that this is a unique moment to provide clinicians with guidance on how to avoid stigma and bias in our language as part of larger efforts to promote health equity. We performed an exhaustive scoping review of the gray and academic literature on stigmatizing medical language. We used thematic analysis and concept mapping to organize the findings into core principles for use in clinical practice. We compiled a list of terms to avoid and seven strategies to promote non-judgmental health record keeping: (1) use person-first language, (2) eliminate pejorative terms, (3) make communication inclusive, (4) avoid labels, (5) stop weaponizing quotes, (6) avoid blaming patients, and (7) abandon the practice of leading with social identifiers. While we offer guidance clinicians can use to promote equity through language on an individual level, health inequities are structural and demand simultaneous systems and policy change. By improving our language, we can disrupt the harmful narratives that allow health disparities to persist.

CLINICAL VIGNETTE

CS is a 58-year-old patient with diabetes and hypertension. He presented to the emergency department (ED) and was found to have cellulitis. He was prescribed doxycycline and discharged. He had two recent ED visits for hyperglycemia because he was unable to secure an outpatient clinic appointment. He has mobility challenges due to neuropathy and lack of reliable transportation. The medication ordered was not on the formulary for his insurance and he could not afford to pay cash for it, so 2 days passed before he obtained an alternative. He returned to the hospital when the cellulitis worsened, as instructed by the physician. He declined admission for intravenous antibiotics because he was concerned about missing work and about his teenage grandson who would be left home alone. The next day, he logged into his electronic health record (EHR) and read the following note: “The patient is a noncompliant diabetic who failed oral antibiotics. He claimed the pain is ‘so bad’ but adamantly refused admission. He’s a frequent flyer so he’ll likely bounce back when it worsens. He signed out against medical advice.”

INTRODUCTION

As healthcare workers, the words we choose in our clinical communication frame narratives about patients and their experiences. Through our language, we transmit messages to each other and to our patients about identity, agency, and blame. At its best, our language can humanize, empower, and build trust, and at its worst, it can exacerbate disparities for marginalized groups.

The conversation around stigmatizing language in medicine began in 1974 with the person-first language movement. Advocates championed terminology that avoids reducing people to their medical conditions.¹ Subspeciality and disease-specific organizations have since increasingly advocated for attention to the language used in clinical communication, given the growing quantitative and qualitative evidence that stigmatizing language influences provider judgments^2–4 and patient experiences when navigating healthcare.^5–7 Research participants, for example, demonstrate more negative attitudes toward the terms “addict” and “substance abuser” compared to “person with a substance use disorder.”^{8, 9} Reading stigmatizing language, such as that which casts doubt, portrays the patient negatively, or implies patient responsibility, has been associated with negative provider attitudes and less aggressive pain management of patients with sickle cell disease, for example.¹⁰ The way we label patients also affects how they interface with the healthcare system. The perception of patients with addiction as “difficult” can prompt providers to ignore patients’ symptoms or to refer them to other clinicians rather than prescribing appropriate medications.⁴ Patients who are termed “morbidly obese,”⁵ “alcoholic,”⁶ and “convict”⁷ are less likely to access medical treatment when they perceive stigma from healthcare providers.

Beginning in the spring of 2021, the 21st Century Cures Act mandated patients have access to their own medical records free of charge.¹¹ This legislation provided impetus for the medical community to continue the important conversation about how we speak and write about our patients. Data from prior to this legislation demonstrated that 60% of patients with access viewed their EHR at least one time.¹² In a survey study, the majority of physicians who had used open notes agreed it was a good idea and viewed the change as useful. Both PCPs and specialists indicated that open notes did not change the value of their notes for other providers. Respondents cited that they most frequently changed “language that could be perceived as critical of the patient.”¹³

Since then, bias and stigma continue to be identified in clinical documentation. A large cross-sectional study of admission notes identified stigmatizing language in 2.5% of documentation, and with higher frequency for patients with diabetes.¹⁴ In a large survey study of outpatients with EHR access, 10% of patients reported feeling judged and/or offended by their provider’s language.¹⁵ Patients with poor health or who were unemployed were even more likely to feel this way. Specific themes identified in the EHR and perceived by patients as negative included errors/surprises, labeling, and disrespect. Another recent qualitative study of 600 outpatient notes described several common themes of both negative and positive language in physician notes and called for increased attention to language choice to mitigate bias. The five identified patterns of negative language were disbelief, disapproval, stereotyping, references to uncooperativeness, and unilateral decision-making.¹⁶ Additionally, several recent studies have shown that stigmatizing or biased language is used more frequently in the medical records of non-Hispanic Black patients, raising concern for the potential for language to exacerbate health disparities.^{14, 17, 18}

Subspecialty and patient advocacy groups have developed educational materials for clinicians about stigmatizing language (see Table 1). For instance, the American Diabetes Association (ADA)²⁰ convened a task force of diabetes educators who created a consensus guideline of five general recommendations, including language that is neutral, stigma free, respectful, collaborative, and patient centered. The organization also provides specific alternative language to use in place of common phrases with negative connotations. The National Institute on Drug Abuse (NIDA)³¹ compiled a guide of terms to use, terms to avoid, and rationale; it is available on their website entitled “Words Matter: Preferred Language for Talking About Addiction,” which cites the relevant literature and provides suggested language for talking about people, substances, recovery, and treatment. Obesity UK, a charity dedicated to supporting people living with obesity, created the “Language Matters: Obesity” guide³⁵ based on feedback from clinicians, dieticians, clinical psychologists, and conversation analysts. It includes direct quotes from service users about their experiences and preferences. Although specific guides such as these exist, no comprehensive resource assembles these recommendations for generalists. Given this gap, we sought to review the academic and gray literature, identify themes, and provide a set of guiding principles for clinicians in the USA to reduce stigmatizing language and bias in the medical record.

Table 1.

Sample of Stigmatizing Language Education Campaigns

Organization or institution	Title of campaign, initiative, or resource	Topic
American Academy of Addiction Psychiatry19	Changing Language to Change Care: Stigma and Substance Use Disorder	Substance use disorder
American Diabetes Association20	The Use of Language in Diabetes Care and Education	Diabetes
American Psychiatric Association + SAMHSA21	Words Matter: the Importance of Person First Language	Mental health
American Psychological Association22	Bias-Free Language	Disability, gender, sexual orientation, etc.
Boston Medical Center23	Words Matter	Substance use disorder
Canadian Centre on Substance Use and Addiction24	Overcoming Stigma Through Language	Substance use disorder
Canadian Public Health Association25	Language Matters	Substance use disorder, sexuality, STBBIs
Colorado Office of Behavioral Health26	Lift The Label	Opioid use disorder
Dartmouth-Hitchcock Health27	Substance Use and Mental Health Initiative	Mental Health, substance use disorder
Indiana Recovery Council28	Do You Speak Stigma?	Mental health, substance use disorder
Michigan.gov29	End the Stigma	Opioid use disorder
National Health Service England30	Language Matters	Diabetes
National Institute on Drug Abuse31	Words Matter	Substance use disorder
Network of Alcohol and other Drug Agencies32	Language Matters	Substance use disorder
North Carolina Department of Health and Human Services33	Stop the Stigma	Substance use disorder
Northeastern University School of Law - Health In Justice Action Lab34	Changing the Narrative	Substance use disorder
Obesity Canada35	Language Matters	Obesity
Ohio Department of Mental Health and Addiction Services36	Words Matter	Substance use disorder
Prevention Solutions37	Words Matter	Substance use disorder
Recovery Research Institute38	Addictionary	Substance use disorder
Shatterproof39	Stigma-Reducing Language	Substance use disorder

METHODS

Two authors (MH, AR) searched both Pubmed and Google with key terms to identify relevant peer-reviewed literature, anti-stigma campaigns, and guidance from public health and advocacy organizations. For academic literature on the use of stigmatizing language in healthcare, we searched PubMed between September 2020 and February 2022 with the following key terms: “bias,” “healthcare,” “intervention,” “medical record,” “mental illness,” “patient outcome,” “person first language,” “stigma,” “stigmatizing language,” and “substance use disorder.” Additionally, we searched Google over the same period to identify institutional and organizational campaigns, initiatives, and educational resources promoting the use of non-stigmatizing language using the key search terms above and, additionally, the following: “words matter,” “language matters,” “campaign,” and “initiative.” The combined search yielded 657 results, of which 54 met inclusion criteria (see Fig. 1). We included studies that focused on clinician language in written documentation about patients. We excluded sources focused exclusively on stigma and bias in interpersonal communication, research, community engagement, or other health and science communication.

Figure 1.

Flow diagram of search results.

Given that content presentation in our sources was highly heterogeneous and included qualitative and quantitative research studies, white papers, and websites, we chose not to code textual data verbatim. Instead, in the first pass of data extraction, two members of the research team (AR and MH) both independently read the full text of papers from the academic literature and guidelines, websites, and white papers from the gray literature. During this process, they independently took notes on each source focused on key concepts in written language preferences among patients. Then, in a series of three discussions, the two researchers (AR and MH) met to reconcile these notes and develop descriptive paragraphs for each source.

The two researchers then grouped similar sources together based on (a) specific words and phrases to avoid and (b) strategies to make language more neutral. They eliminated strategies that did not have consensus (defined as being mentioned in three or more sources). For strategies and terms where the two researchers could not agree, a third researcher (KK) acted as a tiebreaker by reading the related full papers or gray literature documents to decide whether the concept had enough salience in the literature to be included.

Finally, in three subsequent theory-generating discussions, the wider team used the process of concept mapping to arrange the included strategies into broader principles.⁴⁰ The goal here was to abstract from the data principles with validity outside of the specific context for which the recommendations were intended. Wherever possible, specific strategies were distilled into guiding principles that could be applied more broadly to writing about patients in the electronic health record. For example, literature preferred the phrasing “person with substance use disorder” over “addict” and person with diabetes over “diabetic.” These strategies were therefore distilled into the guiding principle of using person-first language. As we distilled these broader concepts, we built a table that grouped sources that corresponded to each theme (see Table 2). In this process, the strategies with consensus that pertained to the use of a single word or term were compiled into a list of pejorative terms to avoid (see Table 3).

Table 2.

Stigmatizing Language Guide

Guiding principle	Rationale	Sample stigmatizing phrasing	Alternate language
Use person-first language.7–9, 22, 30, 41–43	• People are more than their conditions. • Terms should acknowledge patients’ humanity first. • Avoid using adjectives as nouns. • Be aware that a few specific communities prefer identity-first language (ex. “Autistic person”).	Diabetic	Person with diabetes
		Other examples: schizophrenic, sickler, dialysis player, vasculopath	Person with schizophrenia, sickle cell disease, ESRD, PVD
Eliminate pejorative terms.2, 41, 44–48	• Loaded terms propagate negative attitudes and influence care. • Criminal-legal terms and combat language should generally be avoided.	The drug screen is dirty.	The drug screen is positive.
		Other examples: Addict Drug user	Person who uses drugs/injects drugs (PWUD/PWID) Person with substance use disorder (SUD) Person with opioid use disorder (OUD)
		Convict, inmate, felon	Person who has been incarcerated
		Relapse	Return to use, recurrence of drug use
		Track marks	Injection-related wounds
		Risky behaviors	Condomless sex, multiple partners
		Cancer sufferer/victim; battle/war with cancer	Person living with cancer
Choose inclusive language.30, 49, 50	• Utilize the person’s correct pronouns. When in doubt, use neutral pronouns. • Transgender is an adjective, not a noun. • Regarding disability, acknowledge circumstances but avoid disempowering narratives.	Wife, girlfriend	Spouse, partner, significant other
		Gay lifestyle, preference for male partners	Sexual orientation
		Wheelchair bound	Uses a wheelchair
		Afflicted with/suffers from/victim of	Lives with/has
Avoid labels.16, 43, 51–53	• Generalizations are not helpful and may be harmful. • Labels can be propagated through successive notes without exploration of details. • Avoid adversarial or negative framing.	Patient is noncompliant.	He is not taking medications.
		Patient is a poor historian.	I had trouble taking a history due to x.
		Difficult, manipulative, unfortunate patient
Do not weaponize quotations.10, 16, 30	• Avoid undermining patients’ experiences. Putting facts in quotations infers a lack of credibility. • Quotes should not be used for humor. • Recognize times when quotations can be useful—such as when describing the quality of a symptom.	The patient has “stress at home.”	He has multiple stressors including….
		The patient reports the pain is “so bad.”	She has severe pain.
Do not lead with race, ethnicity, language, socioeconomic status, and other social identifiers.10, 16, 22, 54–58	• Leading with identifiers in the one-liner propagates bias and reifies biologic race. • Be intentional and specific when using social identifiers. • Beware assumptions—self-report of identifiers is best.	65-year-old African American male presents with chest pain.	65-year-old presents with chest pain.
Avoid language that attributes responsibility to patients for their conditions.16, 30, 51, 59	• Language should not ascribe blame or judgment. • When possible, identify systems or provider-level barriers that can be mitigated.	The patient failed outpatient treatment. He is refusing to be admitted.	The antibiotics did not work for this infection. He declines admission.
		Patient is refusing to wear oxygen.	She is not tolerating oxygen.
Avoid verbs that undermine the patient’s experience.10, 16, 18, 20	• The history of present illness is a subjective experience. • Poor verb choice compromises the credibility of the patient’s narrative. • Some verbs imply the narrator is unreliable.	Patient claims/alleges 10/10 pain.	He is in/was in/had 10/10 pain.

Table 3.

Non-Exhaustive List of Pejorative Terms to Avoid in Clinical Communication

Addict Alien Alcoholic Amputee Cripple Committed (as in suicide) Compliant/compliance Convict/ex-convict Dirty/clean Drinker Drug abuse/intravenous drug abuse (IVDA) Druggie Drug user Drunk Felon Frequent flyer Habit/drug habit Handicapped Illegal immigrant Illicit/street drugs Inmate Intravenous drug user Junkie Medication seeking/drug seeking Mental retardation Narcotic Noncompliant Offender Poor historian Prisoner Promiscuous Prostitute/prostitution Risky/unsafe Shooter Sickler Sufferer Transvestite Vasculopath Victim

While compiling the recommendations, the team sought feedback from content experts from the following areas to identify additional useful resources that may have been missed: addiction medicine, harm reduction, and LGBTQ health. Earlier versions of the guideline were also presented to a Patient and Family Advisory Council (PFAC), institutional committees, and community-based organizations for feedback (see Acknowledgements).

Statement of Reflexivity

This research was conducted by a team that includes an emergency physician, an internal medicine physician, and a medical student who work clinically in two major northeastern US cities. All authors are from privileged backgrounds and do not have significant lived experience with the stigmatized medical conditions covered in this paper. Each of the authors have spent significant time working with historically marginalized communities domestically and/or globally, including work with patients with substance use disorders, mental illness, and HIV. Throughout the analysis, the authors reflected upon their positionality and discussed their potential biases and other limitations conferred by a lack of lived experience with the conditions being studied. The authors sought input from content experts and community representatives when able to mitigate their limited positionality.

RESULTS

This review aggregates academic literature and anti-stigma campaigns to develop a set of guiding principles for medical documentation. Table 1 includes a list of the 21 sources from the gray literature used to create the guideline. Overall, the academic and gray literature recommend that providers use person-first language, eliminate pejorative terms, make communication inclusive, avoid labels, stop weaponizing quotes, avoid blaming patients, and abandon the practice of leading with social identifiers.

Person-First Language

Person-first language stands out as a key principle across the guidance.^{8, 41, 44} To humanize the medical record, sources stress the importance of separating individuals from their conditions. For example, “person with diabetes” is recommended by both academic authors and advocacy groups over “diabetic.”³⁰ Overall, language that respects patients’ autonomy is recommended over victimizing language that diminishes their agency.³⁰ While person-first terminology is a general rule, context matters. Depending on the community, and especially regarding disability, guidance is still evolving. Person-first language (“person with paraplegia”), identity-first language (“autistic person”), or both may be acceptable depending upon the group in question. Additionally, some patients may use terms such as “addict” for self-identification, but available resources advise that this language should not be used by people outside that community.⁴⁵ As noted in multiple language guides from the gray literature, clinicians should defer to patient preferences when feasible and appropriate.²⁰ A patient-provider discussion supports the underlying goal of respectful, collaborative communication.

Pejorative Terms

According to the sources identified, pejorative terms like “victim,” “frequent flyer,” or “offender” exacerbate preexisting power differentials between patients and providers by positioning the patient as inferior. The literature calls for alternatives that are judgment free, inclusive, and values based. A non-exhaustive list of pejorative terms is included in Table 3. Many of the terms are from criminal-legal or military language. Other terms represent outdated or inaccurate understanding of disease.

Inclusive Language

The use of inclusive language across a variety of identities and conditions was stressed throughout the sources. This is especially important for patients who are members of groups that have been marginalized and have resultant health disparities, such as LGBTQ people or people with disabilities. A theme across the guidance was for healthcare providers to default to neutral terminology and avoid disempowering language choice, for example, “wheelchair bound.”²²

Labeling

Labeling of patients was a common finding in clinical communication. Labels such as “noncompliant,” “poor historian,” and “manipulative” can transmit judgment and bias to other care team members and should be avoided. Instead, literature encourages clinicians to identify specific actionable barriers when possible, including individual or social factors or systems-level challenges.³⁵

Use of Quotations

Several sources identified the misuse of patient quotations as a practice that can propagate stigma and bias.^{10, 16} For example, quotes that use stereotypical vernacular elicit negative attitudes by unnecessarily referencing socioeconomic status or racial/ethnic identifiers. In addition, overuse of quotations to report otherwise subjective information (as in “patient reports the pain is ‘so bad’”) can cast doubt on the reporter’s reliability.

Social Identifiers

Leading with social identifiers in the medical record can reinforce systemic biases. Social identifiers that function as markers of marginalization include race, ethnicity, language, and socioeconomic status. Equity scholars stress that including patient race without clear reason and without a discussion of racism reifies biologic race and propagates harm.^{54, 55} As a population health concept, race may be useful to target interventions⁵⁶; however, at the individual patient level, its reporting is more likely to be harmful, especially if it is assumed by providers instead of self-reported.^{57, 58} Other potential markers of marginalization like ethnicity, language, and socioeconomic status can bias providers, especially when they enter the “one-liner.” Social identifiers are complex and intersectional, and using them as shortcuts in clinical communication is not recommended.¹⁰ Providers are encouraged to take time and care to use relevant and appropriately contextualized descriptors that accurately reflect the patient’s experience in the social history portion of the medical chart.

Blame

Throughout the literature, avoiding patient blame emerged as a strong theme. One identified actionable strategy to avoid blame is to improve language around choice, for example, by using “declines” or “chooses” as opposed to “refuses.”^{30, 51} Similarly, the concepts of treatment “compliance” and “adherence” are insufficient.⁵⁹ As an alternative, clinicians can note specific barriers to the treatment plan, such as transportation, cost, or polypharmacy.

Verb Choice

Finally, multiple sources cited verb choice as an important consideration in the pursuit of non-stigmatizing communication. Poor verb choice can influence provider attitudes negatively by implying patient blame, casting doubt (as in “patient claims”), or referencing uncooperativeness.¹⁶

DISCUSSION

Human disease is a sociologic phenomenon as much as a biologic one. Language serves as an undervalued tool as we work to improve health. As our understanding of illness deepens and broadens over time, our language evolves to reflect what we know about the complex human condition. Meaning can also vary depending on context; certain communities have not reached consensus about preferred terminology. Improving our clinical communication to ensure respect, compassion, and justice requires clinicians to remain open to continued learning from our patients and their experiences. We presume that providers agree about the value of using neutral, shared language that best promotes patient dignity and respect. We made these guidelines not to be static or definitive, but rather to, as best as possible, provide a non-binding and practical guide for general practitioners who care for patients with diverse identities, experiences, and medical conditions.

Our overarching recommendations stress the avoidance of patient blame. We recognize that patients make individual choices that impact their health and have some individual accountability. However, emerging literature shows that patients from the most marginalized groups are more likely to have stigmatizing language in their medical charts.^16–18 While patients do have agency, for these patients, such individualized forms of power are typically overshadowed by structural factors that are out of a patient’s control. To overemphasize patient responsibility risks eliding the role of structural racism, sexism, ableism, and other forms of oppression on patients' lived experiences. We can avoid compounding this harm for patients who may be particularly vulnerable to this language. One actionable way to do this is by avoiding negative judgments in our clinical documentation. Emerging research suggests the potential relational benefits to both providers and patients when this is thoughtfully done.^{60, 61}

Overall, we found that the academic literature in this field is nascent. We communicate with our patients and one another in both direct and indirect ways in clinical practice, which makes it challenging to identify and quantify these forces at work. Most relevant academic research examines the effect of language on provider attitudes,^{2–4, 8, 9} patient and provider experience,^{5–7, 13, 15} and clinical decision-making¹⁰ rather than clinical outcomes. Additional research to fill these gaps would add to efforts for providers to become more patient centered. We focused on clinical documentation, but research that examines stigma and bias in interpersonal communication, research, community engagement, or other health and science communication is also needed.

Since much of what we know about stigmatizing language comes from advocacy groups that speak on behalf of people with lived experience rather than researchers, we conducted a narrative rather than a systematic review to broaden the scope to include both academic and gray literature pertinent to this topic. The objective was to create a comprehensive clinical reference, rather than to answer a pointed research question (as would be the case with a systematic review). While this introduces bias regarding article selection, we have made efforts to ensure that our search accounted for sources that were developed with stakeholder input. Future research could develop a more robust consensus guideline via the Delphi method⁶² and ensure broader representation from healthcare workers, patients, and community members using participatory action-based research.⁶³ To guide our decisions and prevent harm, we must simultaneously build upon the emerging quantitative data, seek out rich qualitative evidence, responsibly partner with community representatives, respect the expertise of those with lived experience, and continually return to the ethical foundations of our field. Above all, in clinical communication, we call for reflection on the core values of medicine: compassion, respect, integrity, ethical practice, and justice.⁶⁴

CONCLUSIONS

Although this review emphasizes the individual steps we can take as clinicians to address stigmatizing language, health disparities are rooted in structural inequity. Racism, poverty, and historical discrimination drive downstream health outcomes through harmful policies and practices⁶⁵ and biased language is one way damaging narratives about patients and communities are reinforced, allowing oppression to persist. Improving the language we use is one important and actionable step to prevent harm, but to address health inequities, there must be concurrent efforts to uncover and address structural determinants of health.

Despite increased attention to existing disparities, tragic differences in health outcomes for marginalized communities persist. Consideration of the social construction of illness offers new insights and a path forward.⁶⁶ Language is one powerful way we make meaning of illness and it in turn drives our societal response. We must be aware of the way our language demonstrates the limits of our medical knowledge and in turn reflects and reproduces social inequality.¹⁹ On the positive side, language can be leveraged to combat illness by reframing our understanding of disease and its root causes. However, stigma and bias are intertwined, so that when structural causes go unaddressed, individual patients bear the burden in a cycle of harm.⁶⁷ For example, perceived blame as communicated to and about patients influences everything from an individual person’s likelihood of accessing care to the funding of a particular health-related initiative. Healthcare providers can intervene at the individual, biomedical level, striving to fill gaps and hold patients accountable for their circumstances. Or, alternatively, we can aim for something more elusive but potentially more powerful: to change the dominant narrative about illness itself.

Declarations

Conflict of Interest

All authors had no conflicts of interest to declare.