Patients can now open the clinical note. Studies of what clinicians write, who is described in judgmental terms, how patients react when they read it, and who logs in at all show that the benefit of reading and the risk of what is written fall on overlapping groups
Open Note has explained what patients can now see in their records. This installment turns to the document itself: the clinician’s note, written to colleagues, which patients can increasingly read. It asks what that note contains. Studies of notes in several U.S. health systems have measured how often they describe patients with judgmental or doubting words, whether that language differs by race and ethnicity, how patients react when they read it, and who is using the portal to find out. The orienting fact is short. Federal policy has expanded patients’ rapid access to clinical notes, as two of the papers cited below describe. This article is about what is on the page once the door is open.
The angle is new for the series because it puts the author of the note under the same lens as the reader. A note is usually described as a patient’s record. In practice it is also a professional’s account of a patient, shaped by habits of language that were developed for colleagues. When patients can read it, the account becomes accountable to the person it describes. MB treats the studies here as evidence about documentation habits across many notes. They do not describe any one clinician, and MB has not reviewed any individual’s chart.
|
AN ILLUSTRATIVE SCENARIO (NOT A
REAL PERSON) This is a composite
built to organize questions. It is not a real patient or clinic, and nobody
MB interviewed wrote it. A patient logs in
after a visit to check the follow-up plan. The note records the visit in
clinical shorthand. One line says the patient “denies” missing doses. Another
says the patient was “noncompliant” with a prior plan, without mentioning
that the pharmacy had been out of stock. Nothing in the note is untrue in a
narrow sense, and the clinician may have written it in ten seconds between
visits. The patient reads it slowly. The questions this
article asks are whether such wording is common, whether it is distributed
evenly across patients, and what reading it does to the patient’s trust and
willingness to keep logging in. |
What is stigmatizing language in a note?
Researchers do not agree on one list of words, so each study defines its own. A 2026 qualitative analysis of 696 notes from a large U.S. academic health system, drawn using natural language processing, built a taxonomy of how notes discredit patients. It identified six categories: patients portrayed as socially undesirable, difficult to interact with, incompetent, manipulative, noncompliant, and not credible. The authors found that stigma was conveyed by more than keywords. Intensifiers, wording that marks distance from the patient’s account, and euphemisms that cast the clinician as the neutral party also carried it. These matters because studies that count words will miss some of the language and will count some words that are used neutrally. Keep that in mind for every number below.
Who is described this way: the measured differences
A 2022 cross-sectional study at one large urban academic medical center analyzed 48,651 hospital admission notes about 29,783 patients written by 1,932 clinicians in 2018. It found stigmatizing language in 2.5% of all notes. In notes about non-Hispanic Black patients compared with non-Hispanic White patients, the probability of containing stigmatizing language was 0.67 percentage points greater (95% CI 0.15 to 1.18), with similar differences in notes about patients with diabetes, substance use disorder and chronic pain. Notes about patients with diabetes contained diagnosis-specific stigmatizing language 6.9% of the time. Notes by clinicians earlier in training contained more of it. The absolute gap is small, and the authors’ measure was a defined word list, so it is a floor on what a reader might experience as judgment.
A second study used machine learning on 40,113 history and physical notes from 18,459 patients at another urban academic center between January 2019 and October 2020. It searched for sentences containing a negative descriptor of the patient or the patient’s behavior, such as “resistant” or “noncompliant.” After adjusting for sociodemographic and health characteristics, Black patients had 2.54 times the odds of having at least one negative descriptor compared with White patients. A 2026 follow-up from members of that research group examined notes about patients with sickle cell disease, a condition that mostly affects people who identify as Black and that is often accompanied by pain treated with opioids. The authors coded seven negative descriptors. Fifteen percent of notes about patients with sickle cell disease contained one, and the odds were 2.46 times as high as for Black patients without the disease (95% CI 1.62 to 3.73), though not different from patients with opioid use disorder.The result shows that race and diagnosis can combine, and the study cannot separate their effects beyond what the comparison groups allow.
Language that questions a patient’s word is a related target. A 2024 study of 54,936 admission notes at an academic health system found that 71.0% contained at least one term casting doubt on the patient’s clinical history, a figure that reflects how broad the study’s definition of doubt was. Notes about non-Hispanic Black patients had 21% higher adjusted odds of containing such a term than notes about non-Hispanic White patients (adjusted odds ratio 1.21; 95% CI 1.14 to 1.28). Notes about other racial and ethnic minoritized groups had similar odds to notes about White patients. A much larger analysis in a mid-Atlantic health system examined 13,065,081 notes about 1,537,587 patients. It found that only 0.82% of notes contained terms undermining a patient’s credibility, and 0.26% contained terms supporting it. Notes about Black patients had higher odds of undermining terms (adjusted odds ratio 1.29) and lower odds of supporting terms (0.82). Notes about Hispanic or Latino patients were similar to those about White patients, and notes about Asian patients had lower odds of undermining terms (0.85) and higher odds of supporting terms (1.30).
That last set of results is a useful reminder not to treat the findings as a single line. Differences ran in different directions for different groups, and the studies were each conducted at one or two health systems. They show documentation patterns in the places they looked. They do not tell a reader about any particular clinic, and they do not identify the reason a clinician chose a word.
Other characteristics beyond race show up in the same measurement. A 2026 retrospective cohort of 1,433 pregnancies complicated by diabetes found stigmatizing terms such as “noncompliant” and “refused” in the records of 128 (8.9%). Birthing individuals of color had about four times the odds of having such a term documented (adjusted odds ratio 3.87), people with public insurance had nearly five times the odds (4.95), and people who spoke a language other than English or Spanish had three times the odds (3.18).The groups are the authors’ categories, drawn from one academic-community database, and the study could not examine why the terms were used.
|
Study
(setting) |
What
was counted |
Difference
reported for Black patients vs White patients |
|
Himmelstein
2022 (admission notes, one academic center)[1] |
Stigmatizing
words, defined list |
0.67
percentage points greater probability of stigmatizing language |
|
Sun 2022
(history and physical notes, one academic center)[2] |
Negative
descriptors, e.g. “noncompliant” |
2.54 times
the odds of at least one descriptor |
|
Lee 2024
(admission notes, one academic health system)[3] |
Words casting
doubt on history |
Adjusted odds
ratio 1.21 |
|
Beach 2025
(13 million notes, mid-Atlantic system)[4] |
Terms
undermining or supporting credibility |
Undermining:
adjusted odds ratio 1.29; supporting: 0.82 |
What happens when patients read it
The OpenNotes program has surveyed patients for years. One of its large surveys invited 136,815 portal users at health systems in Boston, Seattle and rural Pennsylvania, and 28,782 responded (21.7% of those invited). Among the 22,947 respondents who had read at least one note, 72.62% rated reading as very important for taking care of their health, 3.3% were very confused and 4.83% were more worried after reading. Less educated, nonwhite, older and Hispanic patients, and people who usually did not speak English at home, were the most likely to report major benefits.The respondents were mostly white (76.94%), were portal users who chose to answer, and represent a minority of those invited, so they are not a random sample of all patients.
A separate analysis of 22,959 respondents from the same research effort asked who felt judged or offended. About 10.5% (2,411 people) said so. Patients who reported poor health, unemployment, or inability to work were more likely to feel judged or offended. Among those who explained why, the themes were errors and surprises, labeling, and disrespect.[8] The authors link these reactions to a risk to trust, but the survey asked about feelings after reading. It did not measure whether trust later changed or whether people stopped reading.
Studies from Sweden and Finland, where national portals show records to patients, find higher rates of offense among people who have received mental health care. In a Swedish survey of 12,334 portal users, 37.78% of respondents with mental health care experience reported an offense in their record compared with 17.56% of other respondents.[11] A Finnish survey of 4,681 respondents found that 25.4% of patients who had mental health care felt offended, compared with 9.9% to 11.6% of patients in the cancer, other-conditions and no-care groups. Errors, disrespectful language and unnecessary information were the most frequent reasons.[12] These are other countries and other systems, and the results do not transfer directly to U.S. practice. They matter here because they show the same pattern: the people whose conditions already carry stigma are the ones likelier to report being hurt by what they read.
Qualitative work gives the reaction its texture. A 2026 study of 370 women with a history of pregnancy and substance use disorder, recruited through an online platform and shown examples of clinical language, reported that wording such as “denies” and quotations placed in quotation marks could be perceived as stigmatizing and decrease trust in providers. Participants described expecting stigma from future clinicians who would read the same note, and some said it could affect their care decisions for themselves and their babies. They also wanted nonstigmatizing language and acknowledgement of recovery.[10] The study is qualitative, relies on a volunteer sample, and shows perceptions rather than behavior. It is useful because it names specific phrases and explains how a note can be read as a prediction of how the next clinician will treat the person.
Who can log in to read it
A transparency policy only works for people who can use the portal. A 2026 cross-sectional study of 72,417 neurology patients at two academic medical centers in Washington, DC found that 64.7% had activated a portal account. Activation was 76.1% among non-Hispanic White patients, 57% among non-Hispanic Black patients, 55% among Hispanic patients and 57.6% among non-Hispanic Asian patients. After adjustment, the odds of activation were lower for Black (adjusted odds ratio 0.46), Hispanic (0.34) and Asian (0.47) patients than for White patients. Activation varied across city wards from 48% to 82%, and ward-level activation correlated strongly with educational attainment, broadband access and median income.[13] A pediatric orthopaedic practice found that families of Black (odds ratio 0.54), Hispanic or Latino (0.69) and Asian (0.64) patients, and families whose primary language was Spanish (0.21), had lower odds of sending a portal message than White, non-Hispanic and English-speaking families.[14] A needs assessment of 204 patients at a New York City primary care clinic serving mostly Latino and older adults found that about 43% completed a task of reading a physician note correctly in the portal, and most respondents wanted to learn more.[15]
These are three specific settings, one specialty each, with different measures of use. They support a narrow point. Portal use differs by race, ethnicity, language, age and neighborhood resources in the places studied, and the differences track the factors that shape internet access and digital skill as much as they track interest.
Putting the pieces together
Here is the pattern in plain terms. In several health systems, notes about Black patients more often contained stigmatizing, negative or doubting language than notes about White patients, though not in every comparison and not for every group. When patients read their notes, most reported benefit, and the groups most likely to report major benefit included nonwhite and Hispanic patients, but a sizable minority felt judged, and that reaction was more common among those in poor health. And in the places where portal use was measured, activation was lower for Black, Hispanic and Asian patients than for White patients.
The studies were done in different places and on different people, so they cannot be stacked into a claim about one patient’s experience. None of the retrieved studies measured whether a patient who reads a stigmatizing note logs in less afterward, or whether that differs by race. That is the missing study. The plausible concern, that the patients with the most to gain from reading may be among the least likely to be in the portal, is a hypothesis built from separate findings. MB presents it as a question for health systems to answer with their own data.
|
WHAT THE EVIDENCE CAN AND CANNOT
TELL US It can tell us that
stigmatizing, negative and doubting language appears in clinical notes in
several health systems, and that studies at those systems found it more often
in notes about Black patients than White patients.[1][2][3][4] It cannot tell us how
often it appears in any other system, why a clinician chose a given word, or
whether any individual note was written with bias. The studies use word lists
and algorithms, and the taxonomy study shows these miss some language.[7] It can tell us that
most surveyed note readers reported benefit and that about one in ten
reported feeling judged or offended.[8][9] It cannot tell us
whether reading a stigmatizing note changes trust or later portal use for
patients in the U.S., or whether that differs by race. The qualitative
evidence suggests it can, but it measures perceptions.[10] It can tell us that
portal activation and messaging were lower for several groups of patients in
the settings studied.[13][14] It cannot tell us the
national picture, since each portal study covers one specialty and region. |
If you read your own note
Reading a note does not require a medical vocabulary, and a line that surprises you is information, not an accusation. The questions below are for the next visit or a portal message, and none of them assumes a clinician acted in bad faith.
|
QUESTIONS AND STEPS FOR READERS Check the facts
first: medicines, doses, dates, allergies, and what you said happened. Errors
and surprises were a leading reason patients felt judged in the survey.[8] If a word or phrase
bothers you, ask the clinician what it meant and whether it can be reworded.
Ask the practice what its process is for adding your comments or requesting a
correction. If a note describes
you as “noncompliant,” “refused” or “denies,” ask what the clinician was
trying to record, and tell them the barrier if one exists, such as cost,
stock, work hours or language. If you cannot log in,
ask the practice for help with activation, in your language if needed. A
family member can ask the practice what proxy access is available. Ask whether the
practice has a way to flag or discuss wording that patients find hurtful. |
What MB will ask next
MB will ask health systems whether they measure stigmatizing language in their own notes, whether they track portal activation by race, ethnicity, language and neighborhood, and whether they track what happens to portal use after a patient reports a hurtful note. It will ask what training clinicians receive on wording, and whether a patient has a clear route to respond to a note in writing. These are design questions about a record that now has two audiences.
General information, not individual medical advice. For emergencies, seek immediate help. The conversation continues.
■ ■ ■
SOURCES (PEER-REVIEWED WORK RETRIEVED FROM PUBMED; DOI LINKS INCLUDED)
[1] Himmelstein G, Bates D, Zhou L. Examination of stigmatizing language in the electronic health record. JAMA Netw Open. 2022;5(1):e2144967. DOI: https://doi.org/10.1001/jamanetworkopen.2021.44967 (PMID 35084481). Cross-sectional, 48,651 admission notes, one center.
[2] Sun M, Oliwa T, Peek ME, Tung EL. Negative patient descriptors: documenting racial bias in the electronic health record. Health Aff (Millwood). 2022;41(2):203-211. DOI: https://doi.org/10.1377/hlthaff.2021.01423 (PMID 35044842). 40,113 notes, one center.
[3] Lee CR, Aysola J, Chen X, Addisu E, Klein A, Weissenbacher D, Gonzalez-Hernandez G, Weissman GE. Race and ethnicity and clinician linguistic expressions of doubt in hospital admission notes. JAMA Netw Open. 2024;7(10):e2438550. DOI: https://doi.org/10.1001/jamanetworkopen.2024.38550 (PMID 39401039). Cohort, 54,936 notes.
[4] Beach MC, Harrigian K, Chee B, Ahmad A, Links AR, Zirikly A, Han D, et al. Racial bias in clinician assessment of patient credibility: evidence from electronic health records. PLoS One. 2025;20(8):e0328134. DOI: https://doi.org/10.1371/journal.pone.0328134 (PMID 40802585). 13,065,081 notes, one health system.
[5] Wesevich A, Vangelatos A, Sun M, Tung EL, Peek ME. Negative descriptors of patients with sickle cell disease in the electronic health record. JAMA Netw Open. 2026;9(4):e266458. DOI: https://doi.org/10.1001/jamanetworkopen.2026.6458 (PMID 41973422). Cross-sectional, single center, 39,871 notes.
[6] Gwan A, Ortiz IB, Tessier KM, Mahr RM, Looby AA, Molleti S, Makori J, et al. Factors associated with healthcare professionals’ use of stigmatizing language in individuals with diabetes in pregnancy. BMC Pregnancy Childbirth. 2026;26(1). DOI: https://doi.org/10.1186/s12884-026-08891-y (PMID 41772513). Retrospective cohort, 1,433 pregnancies.
[7] McArthur A, Ahmad A, Links AR, Warner KR, Drew P, Beach MC, Saha S. How words discredit: a taxonomy of stigmatizing language in the electronic health record. Patient Educ Couns. 2026;147:109520. DOI: https://doi.org/10.1016/j.pec.2026.109520 (PMID 41689951). Qualitative analysis of 696 notes.
[8] Fernández L, Fossa A, Dong Z, Delbanco T, Elmore J, Fitzgerald P, et al. Words matter: what do patients find judgmental or offensive in outpatient notes? J Gen Intern Med. 2021;36(9):2571-2578. DOI: https://doi.org/10.1007/s11606-020-06432-7 (PMID 33528782). Survey, 22,959 respondents; abstract reviewed.
[9] Walker J, Leveille S, Bell S, Chimowitz H, Dong Z, Elmore JG, Fernandez L, et al. OpenNotes after 7 years: patient experiences with ongoing access to their clinicians’ outpatient visit notes. J Med Internet Res. 2019;21(5):e13876. DOI: https://doi.org/10.2196/13876 (PMID 31066717). Survey, 28,782 respondents; abstract reviewed.
[10] Davis ME, Sharma M, Ancker JS, Pathak J, Sharko M. Role of electronic health record in stigma experiences, trust, and care decisions by women with history of pregnancy and substance use disorder: a qualitative study of perceptions of clinical note language. J Am Med Inform Assoc. 2026;33(9):1748-1758. DOI: https://doi.org/10.1093/jamia/ocag096 (PMID 42319960). Qualitative, 370 respondents.
[11] Bärkås A, Kharko A, Blease C, Cajander Å, Johansen Fagerlund A, Huvila I, Johansen MA, et al. Errors, omissions, and offenses in the health record of mental health care patients: results from a nationwide survey in Sweden. J Med Internet Res. 2023;25:e47841. DOI: https://doi.org/10.2196/47841 (PMID 37921861). Survey, 12,334 respondents.
[12] Simola S, Kujala S, Cajander Å, Kharko A, Kane B, Wang B, Åhlfeldt RM, et al. Understanding patient-reported offenses in electronic health records: cross-sectional mixed methods survey. J Med Internet Res. 2026;28:e86178. DOI: https://doi.org/10.2196/86178 (PMID 42133829). Survey via the Finnish national portal, 4,681 respondents.
[13] Streicher N. Patient portal activation disparities by ward, census tract, and zip code in an urban landscape among neurology patients: cross-sectional multi-scale analysis. J Med Internet Res. 2026;28:e98695. DOI: https://doi.org/10.2196/98695 (PMID 42684304). 72,417 patients, two academic centers in Washington, DC.
[14] Shenoy DA, Pean CA, Allen M, Behman A, Lark RK, Catanzano AA. Providing beyond clinic hours: a single-center analysis of patient portal communication in pediatric orthopaedic care. J Pediatr Soc North Am. 2025;14:100288. DOI: https://doi.org/10.1016/j.jposna.2025.100288 (PMID 41492415). Retrospective cohort, one center.
[15] Glickstein J, Idumonyi JO, Ting T, Robles-Aponte I, Ramey M, Fullilove R, et al. Digital health literacy in safety-net primary care: a needs assessment and call to action for health equity. Front Digit Health. 2026;8:1907871. DOI: https://doi.org/10.3389/fdgth.2026.1907871 (PMID 42866528). Cross-sectional needs assessment, 204 patients, one clinic.
