Patient Family Update Prompt

What this prompt does

A patient family update prompt structures clinical facts you supply into a plain-language message a patient's family can follow — what changed, what it means in everyday terms, what happens next, and who to call. It's the email nurses describe as one of the hardest routine writes of the week: a care plan change, a discharge plan, a status update after a procedure, each one going to a worried reader with no medical training, probably on a phone in a hospital corridor.

The difficulty isn't the facts. It's doing three jobs in one message — translating, reassuring without promising, and being complete without being frightening. That's why these emails take twenty careful minutes when the clinical content would take two.

The medical handoff note prompt covers the other direction: the same shift, written for a colleague in clinical shorthand. "Switched to PO antibiotics, day 3 of 5" is a complete sentence to the incoming nurse and gibberish to the patient's son. This prompt writes for the son.

The prompt

Help me draft a plain-language update to a patient's family member. Use only the information I provide below — do not add, infer, or expand any clinical detail.

**Who I'm writing to:** [RELATIONSHIP AND CONTEXT — e.g., "the patient's son, was present at admission, knows the diagnosis"]
**What changed:** [THE FACTS — a medication change, a care plan update, a discharge decision]
**What it means in everyday terms:** [YOUR OWN PLAIN EXPLANATION — polish my wording, never replace it]
**What happens next:** [UPCOMING STEPS AND ROUGH TIMING]
**What the family can do:** [VISITING, ITEMS TO BRING, DECISIONS TO PREPARE FOR — or "nothing needed"]
**Who to call:** [NAME OR ROLE, NUMBER, BEST TIMES]

Rules:
- Plain language throughout. If a clinical term is unavoidable, keep it and add a short everyday explanation in parentheses on first use.
- Lead with the most important point. Worried readers skim.
- No promises about outcomes, recovery, or timing beyond what I stated above.
- Warm, steady, and factual — no euphemisms, no drama.
- Paragraphs of three sentences or fewer.
- If a section I filled in looks incomplete (no timing, no contact number), flag it with [VERIFY] instead of filling the gap yourself.

How to use it

1. De-identify before you paste. Check your institution's policy on AI tools and patient information — many facilities have specific rules about PHI. Initials and room numbers usually suffice for drafting.
2. Write the "what it means" line yourself. This is the one field the model must never author, because the plain-language explanation of a clinical change is itself clinical content. The prompt's job is structure and tone; the meaning is yours.
3. Review every line against the chart, then read the draft aloud once. If a sentence would confuse the family member you're picturing, it will confuse the real one.
4. Check understanding when it matters. AHRQ's teach-back method — asking people to restate in their own words what they need to know or do — is the standard way to confirm a health message landed. A written update that survives teach-back on the phone is a good sign you translated it well.

Example output (de-identified)

Inputs: patient R.S., 81, admitted after a fall, writing to her son. Changes supplied: antibiotic switched from IV to tablets this morning, physical therapy assessment tomorrow, possible discharge Thursday or Friday.

Subject: Update on your mother — good progress today

Hello Mr. S.,

Your mother had a good day. This morning the team switched her antibiotic from an IV drip to tablets, which means the infection is responding and she no longer needs to be connected to a line.

A physical therapist will assess her walking tomorrow morning. That assessment tells us whether she can go straight home or would do better with a short rehabilitation stay first.

If everything stays on track, discharge would be Thursday or Friday. Nothing is confirmed yet — the therapy assessment comes first.

Two things you can do: bring her walking shoes and reading glasses before tomorrow morning, and if you'd like to join the discharge conversation, morning rounds reach her room around 9 a.m. on Thursday.

Questions before then: call the unit at [VERIFY: unit number not provided] and ask for the nurse looking after Room 12. Evenings are usually easier to reach us than mornings.

"Good progress today" sits in the subject line because the rules put the most important point first — a family member scanning a full inbox gets the reassurance before they've even opened the message. And the [VERIFY] flag at the bottom is the prompt refusing to invent a phone number, which is exactly the refusal you want.

Variations

For discharge day

"Format as a short recap that accompanies the official discharge paperwork and never replaces it. Point to the printed instructions for every medication and follow-up detail rather than restating them — this message covers who to call, what warning signs were discussed, and when the follow-up appointment is."

For a phone call instead of an email

"Format as talking points for a five-minute phone call: the lead sentence, three points in order, and a closing line with the callback number. I'll speak from it, not read it."

After a difficult conversation

Serious news belongs to the care team, delivered in person or by phone — this prompt doesn't carry it. What it can write is the follow-up: "The family spoke with the attending physician this afternoon. Write a brief, gentle recap of the practical points only — visiting arrangements, who to call overnight, where to park after 10 p.m. — and do not restate any medical details from that conversation."

Common pitfalls

• Don't paste real names, dates of birth, or identifiable details into public AI tools unless your institution has approved it. Same rule as every clinical use.
• Don't let the model explain the medicine. If a draft comes back with a clinical explanation you didn't supply, cut it — confident, medical-sounding text is precisely what a language model produces when a field is left thin, and in this context that's a patient-safety problem, not a style problem.
• Don't send the first draft. The structure will be right; the judgment calls — what this family already knows, what they're afraid of — are still yours to layer in.
• Try instead: keep the skeleton and add one real observed detail before sending. "She asked about her garden this afternoon" reassures a worried son more than any polished paragraph about clinical stability.

Who uses this prompt

• Bedside nurses: the end-of-shift family email that otherwise eats twenty careful minutes. More in prompts for nurses.
• Charge nurses and case managers: discharge-planning updates where the family needs timing, logistics, and a contact — not a chart summary.
• Advanced practice providers: written recaps after family meetings, so everyone leaves with the same practical points.

ChatGPT, Claude, and Gemini all handle this prompt unmodified. The load-bearing line is the first one — "use only the information I provide" — which is the constraint-first structure Anthropic's prompt engineering guidance treats as foundational: tell the model what it may not do before what it should. In this prompt, that ordering is the whole safety model. The update that lands, in the end, isn't the most complete one — it's the one a frightened reader doesn't have to read twice.