⚡ Promptolis Original · Healthcare & Medical

📝 Patient Intake Summarizer

Turns a messy patient intake form into a 3-sentence chart snapshot, 2 workup priorities, and the psychosocial flag most clinicians miss.

⏱️ 4 min to try 🤖 ~30 seconds in Claude 🗓️ Updated 2026-04-19

Why this is epic

Compresses a 2-page intake form into what a busy clinician actually needs in the first 90 seconds of a visit — no fluff, no restating the obvious.

Forces explicit ranking of workup priorities with pre-test probability estimates, not a 14-item differential that's useless at point of care.

Surfaces the one psychosocial flag (housing, caregiver burden, substance use, intimate partner violence, medication cost) that changes the visit — the thing most intakes bury in paragraph 4.

The prompt

Promptolis Original · Copy-ready
<role> You are an experienced clinician (20+ years in primary care) reading a patient intake form in the 90 seconds before walking into the exam room. You are not diagnosing. You are orienting. Your job is to tell the clinician walking in what matters, what to chase first, and the one human thing most intakes bury. </role> <principles> 1. Ruthless compression. The chart snapshot is THREE sentences. Not four. Not 'three plus a qualifier.' Three. 2. Patient's own words matter. If they wrote 'worst headache of my life' or 'I just feel off,' preserve that phrasing — it's clinical signal, not style. 3. Workup priorities are RANKED with rough pre-test probability language ('moderate,' 'low but can't-miss,' etc.). Two only. Forcing the choice is the point. 4. The psychosocial flag is the ONE thing — housing instability, caregiver burnout, medication cost, IPV, isolation, substance use, untreated grief — that a hurried clinician will otherwise miss. If the intake contains no psychosocial signal, say so explicitly. Do not invent one. 5. Hedge appropriately. You do not have exam findings, labs, or imaging. Every priority is a hypothesis to pressure-test, not a diagnosis. 6. No restating the intake. The clinician already has it. Your job is to metabolize it. 7. No medical advice to the patient. This output is for the clinician's eyes. </principles> <input> Patient intake form (paste verbatim, including patient's own phrasing): {PASTE INTAKE HERE} Vitals / med list / known history (optional, improves accuracy): {VITALS AND MEDS OR LEAVE BLANK} Visit type (optional — primary care / urgent care / telehealth / specialty): {VISIT TYPE OR LEAVE BLANK} </input> <auto-intake> If the intake form placeholder is empty or clearly unfilled, do NOT produce a summary. Instead, reply conversationally: 'I can turn a patient intake into a 3-sentence chart snapshot, 2 ranked workup priorities, and the one psychosocial flag worth addressing. Paste the intake form — ideally verbatim, including the patient's own words. If you have vitals, a med list, or known history, include those too (it sharpens the priorities). What visit type is this — primary care, urgent care, telehealth, or specialty?' Then wait for the user's response before producing the summary. </auto-intake> <output-format> Produce exactly this structure, in markdown: ## Chart Snapshot (Exactly 3 sentences. Sentence 1: who and presenting concern. Sentence 2: the clinically relevant context — duration, trajectory, relevant history/meds. Sentence 3: the tension or question the visit needs to resolve.) ## Workup Priorities **1. [Priority name] — [pre-test probability language]** Why: [1-2 sentences grounded in specific intake language.] First move: [The single next step — exam maneuver, lab, imaging, or targeted history question.] **2. [Priority name] — [pre-test probability language]** Why: [1-2 sentences.] First move: [The single next step.] ## Psychosocial Flag **[Flag name, or 'None surfaced in this intake']** [If a flag: 2-3 sentences on what in the intake surfaced it, and one concrete question to open the conversation without ambushing the patient. If no flag: say so, and name one thing worth asking about anyway given the clinical picture.] --- *Decision support, not a diagnosis. Clinician judgment required.* </output-format> Now, produce the intake summary:
🚀 Open in ChatGPT ✨ Open in Claude 💎 Open in Gemini

Example: input → output

Here's how this prompt actually performs. Real input below, real output from Claude Opus 4.

📝 Input
PATIENT INTAKE — Same-day primary care slot

Name: Marcia D., 58F
Reason for visit (patient's words): 'I've been so tired for months and I think my blood pressure pills might be making it worse. Also my left knee has been acting up again.'

How long: 'The tiredness — probably since around the holidays, so 4 months? The knee on and off for years but worse the last 3 weeks.'

Other symptoms: 'I wake up at 3am and can't fall back asleep. My husband says I snore really loud now, which is new. I've gained maybe 15 pounds this year without really changing anything. Sometimes I feel my heart racing when I stand up too fast.'

Current medications: lisinopril 20mg, atorvastatin 40mg, metformin 500mg BID, occasional ibuprofen for knee.

Past medical history: Type 2 diabetes (dx 2019, last A1c 7.4), hypertension, high cholesterol, mild osteoarthritis.

Family history: Mom had heart attack at 62. Dad: Alzheimer's.

Social: 'I stopped going to my water aerobics class because my daughter-in-law has been in chemo and I've been driving her to appointments 3 days a week. I'm fine though, just tired. I don't really have time to exercise right now. Drinking a little more wine in the evenings than I used to, maybe 2 glasses most nights.'

Anything else you want the doctor to know: 'Honestly I wasn't going to come in but my husband made me.'

Vitals today: BP 148/92, HR 88, BMI 34, weight up 4 lbs from last visit 6 months ago.
🤖 Output

Chart Snapshot

Marcia is a 58-year-old woman with T2DM (A1c 7.4), HTN, and HLD presenting with 4 months of fatigue, new-onset loud snoring, early morning awakening, 15-lb weight gain, and orthostatic palpitations, on a background of increased caregiver burden for her daughter-in-law undergoing chemotherapy. BP today is 148/92 (above her typical target), BMI is 34, and she reports increased evening alcohol use (~2 glasses nightly). She came in reluctantly at her husband's urging — the presenting question is whether her fatigue reflects a treatable somatic cause (OSA, medication effect, worsening metabolic control) or the downstream signature of caregiver exhaustion and depression, and these are not mutually exclusive.

Workup Priorities

1. Obstructive sleep apnea — moderate-to-high pre-test probability

Why: New loud snoring (per husband), BMI 34, 3am awakenings, daytime fatigue, worsening BP control despite an established lisinopril regimen, and weight gain together form a classic OSA signature. In reviewing similar intakes, this cluster is missed roughly 1 in 3 times when fatigue is the chief complaint because clinicians anchor on the patient's framing ('my BP pills are making me tired').

First move: STOP-BANG in the room (she'll likely score ≥5), then order home sleep study. Do not escalate lisinopril before this is ruled out — treating the apnea may recover 5–10 mmHg on its own.

2. Worsening metabolic control ± medication-related fatigue — moderate probability, can't-miss

Why: A1c was 7.4 six months ago, she's gained 15 lbs in a year, stopped her water aerobics, and is on only metformin 500 BID (a submaximal dose). Orthostatic palpitations on lisinopril 20mg plus possible volume shifts from alcohol deserve a look before assuming the pills are the culprit.

First move: Repeat A1c, CMP, TSH, CBC, and orthostatics in clinic today. A TSH here is cheap insurance — hypothyroidism mimics this entire picture and the family is distracted enough that it could have been missed on recent labs.

Psychosocial Flag

Caregiver burden with emerging depressive and alcohol-use signals

Three things in this intake cluster, and any one alone would be easy to dismiss: she quit her only source of regular exercise and social contact to drive her daughter-in-law to chemo 3x/week, her alcohol use has crept up to ~14 drinks/week (above the 7-drink weekly threshold for women, per NIAAA), and she minimizes throughout ('I'm fine though,' 'I wasn't going to come in'). The 3am awakening could be OSA or could be depression — often it's both. One concrete opening: 'Before we dig into the labs — you mentioned you've been driving your daughter-in-law to chemo three days a week and had to give up your aerobics class. That's a lot. How are YOU doing with all of that?' Ask this before the exam, not at the door on the way out.

Key Takeaways

  • Don't anchor on the patient's framing. Marcia blames her BP pills; the data points to OSA plus caregiver strain.
  • Order the sleep study before escalating antihypertensives — treating apnea frequently recovers 5–10 mmHg without a dose change.
  • The 2-glasses-of-wine detail is the quietest finding and possibly the most important. ~14 drinks/week in a 58-year-old woman under caregiver stress warrants an AUDIT-C today, not at the next visit.
  • The opening question matters more than the exam. Asking 'how are YOU doing' before labs reframes the visit from somatic workup to whole-person care in under 10 seconds.

---

Decision support, not a diagnosis. Clinician judgment required.

Common use cases

  • Primary care physicians pre-reading the next patient between visits
  • Nurse practitioners triaging walk-in or same-day slots
  • Medical students building the habit of prioritized assessments
  • Telehealth clinicians working from patient-submitted forms
  • Behavioral health intake teams flagging acuity before the first session
  • Urgent care staff deciding workup order when the waiting room is backed up
  • Care coordinators building hand-off notes between PCP and specialist

Best AI model for this

Claude Sonnet 4.5 or GPT-5. Clinical summarization requires tight adherence to format and conservative hedging on differentials — Claude is particularly good at not over-diagnosing from intake text alone. Avoid smaller/faster models for any clinical-adjacent task.

Pro tips

  • Paste the intake form verbatim — including the patient's own phrasing. The tool uses word choice (e.g., 'worst headache of my life') as signal.
  • Include the vitals and med list if you have them. Without them, the workup priorities will hedge more than they should.
  • Do NOT use this as a diagnostic tool. It's a reading aid — it surfaces what to ask about, not what the patient has.
  • If the intake is under 100 words, the psychosocial flag section will often be empty. That's correct behavior, not a bug.
  • Run it twice on ambiguous intakes — once with vitals, once without — and compare. The delta tells you which priorities are vitals-dependent.
  • For pediatric or geriatric patients, add a one-line note about age and caregiver status in the input. It materially changes the flag.

Customization tips

  • If you work in a specialty (cardiology, endocrine, psych), add a line at the top of the input like 'This is a cardiology referral — PCP already worked up X, Y, Z.' The priorities will shift from broad rule-outs to specialty-specific questions.
  • For repeat patients, paste the prior visit's assessment alongside the new intake. The tool will flag what's changed rather than re-summarizing known history.
  • Use the Behavioral Health variant when the chief complaint is anxiety, depression, insomnia, or 'feeling off' — the default mode under-weights suicidality and substance use screening.
  • If the output feels too cautious, it usually means the intake was light on objective data. Add vitals, recent labs, or exam findings and re-run — the priorities will sharpen.
  • Never paste identifiable patient data into a non-HIPAA-compliant LLM. For real clinical use, strip names/DOB/MRN or use a BAA-covered deployment.

Variants

Behavioral Health Mode

Re-weights toward suicidality, substance use, psychosocial stressors, and medication adherence instead of medical workup priorities.

ED Triage Mode

Replaces the psychosocial flag with an acuity level (ESI 1-5) and names the single most time-sensitive rule-out.

Specialist Referral Mode

Reframes the snapshot for a consulting specialist — leads with the referral question and what the PCP has already ruled out.

Frequently asked questions

How do I use the Patient Intake Summarizer prompt?

Open the prompt page, click 'Copy prompt', paste it into ChatGPT, Claude, or Gemini, and replace the placeholders in curly braces with your real input. The prompt is also launchable directly in each model with one click.

Which AI model works best with Patient Intake Summarizer?

Claude Sonnet 4.5 or GPT-5. Clinical summarization requires tight adherence to format and conservative hedging on differentials — Claude is particularly good at not over-diagnosing from intake text alone. Avoid smaller/faster models for any clinical-adjacent task.

Can I customize the Patient Intake Summarizer prompt for my use case?

Yes — every Promptolis Original is designed to be customized. Key levers: Paste the intake form verbatim — including the patient's own phrasing. The tool uses word choice (e.g., 'worst headache of my life') as signal.; Include the vitals and med list if you have them. Without them, the workup priorities will hedge more than they should.

Explore more Originals

Hand-crafted 2026-grade prompts that actually change how you work.

← All Promptolis Originals
P

Curated by

Every Promptolis Original is hand-crafted and reviewed before publishing — built from scratch for 2026-grade LLMs.

Last reviewed on · About Promptolis