Humanistic Charting is the design discipline of capturing a patient's own account — in their own words — and organizing it into something a clinician can actually use at the bedside. It began by asking thirty-eight people at NYC Health + Hospitals to describe themselves, and then asking whether that person is who their providers see.
Are you feeling seen?
Start with the methodTwo things, and confusing them is the mistake that breaks the effect.
Capture the patient's own account through directed open-ended questions, and organize it into a structured artifact a clinician can use at the point of care.
The discipline is fixed. It does not change between an emergency department and an ICU.
A specific question set built to that discipline for a given clinical context.
Every context gets its own instrument, built to the same standard. Treating one fixed question list as "the tool" for every setting is what breaks it.
The unit cell is the directed open-ended question: the patient answers in their own words, inside a frame that produces analyzable structure — neither a multiple-choice cage nor an unusable free-text blob.The design constraint the whole method rests on
Every instrument inherits the same four-section shape, calibrated per context, so outputs stay comparable across conditions and settings.
Completion sits with the patient, in time they already have. The narrative reaches the clinician before the encounter — the team engages with the output, not the intake.
The patient tells a story. What arrives in the chart is a fixed narrative with clinical signal in it — findable the way a lab value is findable.
The structure is enforced before the AI ever touches the text.
The architecture generalizes without changing its logic. What changes per context is the question library, the calibration of social-reality items, and the cadence — not the underlying method.
The instrument the thirty-eight participants shaped did not stay in a paper. It runs today at NYC Health + Hospitals — eight sections, available in more than sixteen languages, optional from the first question to the last. Fill it out and you are meeting the method exactly as a patient meets it.
The original was shaped by asking thirty-eight people what mattered, and it collected every way they might be asked — three to ten phrasings for each thing worth knowing, with the choosing left to whoever was holding the tablet. Version 2 makes the choice and keeps every alternative one tap away, so a person who does not connect with one wording is offered another instead of a blank.
Both versions do the same essential thing: the answers are mapped onto a standardized, condensed document, because nobody reads a person in single answers — and nobody reads a person in question-and-answer either. The output is a set of pre-written first-person paragraphs with blanks in them, and the answers fill the blanks. It opens "Hello, my name is ___, and I use ___ pronouns" and carries on in that voice to the end. No labels, no field names, no questions on the page.
Version 2 adds one thing the original did not have: a follow-up light on the care team's copy. Six parts, fifty-six questions, all of them skippable but one.
Nothing in the first three parts can ever raise a light. Who a person is, what they value, and how they want to be spoken to are not findings — and a tool that flagged them would have turned being seen into being screened.
Runs entirely in your browser · nothing is sent anywhereUCSF Parnassus Emergency Department · 29 adult patients across 6 clinicians · pre/post design with two-sided paired t-tests · published in JACEP Open, 2025.
| Patient-rated metric (n = 29) | Pre → Post | Significance |
|---|---|---|
| Clinician knew important information about my life | 3.3 → 4.7 of 5 | p < 0.05 — largest gain |
| Spent enough time with me | 3.9 → 4.8 of 5 | p < 0.05 |
| Would work with this clinician again | 4.1 → 4.8 of 5 | p = 0.002 |
| Overall visit rating | 8.3 → 9.3 of 10 | p = 0.013 |
The clinician score is the one worth pausing on. The most common objection to asking patients who they are is that it costs the workforce time it does not have. The clinicians in this pilot rated the method higher than the patients did.
Before the instrument, before the pilot, there was a question asked at eleven NYC Health + Hospitals sites: describe yourself — and then, is that who your providers see? Each participant produced two portraits of the same person. The distance between them is the finding, and the reason the method exists.
Every glowing point is a room where someone was asked who they are. Touch a borough, or a site, to meet the people who answered there.
Point size reflects how many participants were interviewed at that site. One participant's site was not recorded.
Who they are — the mother's spaghetti, the jazz piano, the grandchildren, the faith. Who they believe their providers see — a diagnosis, a compliance status, a body in a bed.The two portraits
Presented exactly as collected. Names were masked at the point of collection. Nothing below has been rewritten, tidied, or shortened for readability — the grammar, the repetitions and the pauses are the data.
The second question is the one that carries the study. Two participants said yes. Thirty-six did not — and then, in a sentence or two, told us exactly what was missing. The dominant cause was not indifference on anyone's part. It was structural: there is no place in the encounter, and no field in the record, where a person's own account is expected to go.
What it costs when a person's own account of their body is not believed — told through the paintings of Hertz Nazaire, who lived with sickle cell disease and died of it at forty-eight.
Enter →The live instrument, the published methodology, the study that produced it, and the thought pieces. Every file downloads directly — nothing is gated.
If a file does not open, your browser may be blocking the PDF viewer — right-click the card and choose "Save link as" instead.