Claude Prompt Library

Claude Prompts for Physicians (Documentation, Communication, Literature)

20 copy-paste prompts

20 copy-paste Claude prompts for physicians: SOAP notes, patient education, differential aids, literature reviews, and practice operations. Always verify against clinical guidelines — Claude assists, doesn't diagnose.

Clinical Documentation

4 prompts

SOAP Note from Encounter Notes

1/20

[Paste encounter notes]. Convert to SOAP format: Subjective (HPI in patient's words, ROS pertinent positives + negatives), Objective (vitals, exam by system, results), Assessment (problem list with reasoning), Plan (per problem, including pt education + follow-up). Stay within my notes — flag anything I should add but didn't document.

Converts notes into SOAP format.

💡

Pro tip: Documentation gaps cost RVU + create medico-legal exposure. Asking Claude to flag missing standard elements (ROS, exam by system, MDM components) catches what you forgot before you sign.

Differential Diagnosis Brainstorm

2/20

[Patient presentation summary]. Generate a structured differential. Output: top 5 diagnoses ranked by likelihood, top 3 "can't miss" diagnoses regardless of likelihood, what additional history/exam/labs would distinguish, red flags requiring immediate workup. I'll integrate with my clinical judgment — this is brainstorm support, not diagnosis.

Brainstorms differential diagnoses.

💡

Pro tip: Use as a "what am I missing?" check, not as primary reasoning. Bias correction tool — catches the obvious diagnosis you anchored away from. Always your call.

Discharge Summary Draft

3/20

[Paste hospital course notes]. Draft a discharge summary: admission diagnosis, hospital course (chronological, key events), procedures, consultations, condition at discharge, discharge medications (with reconciliation), follow-up plan, patient education provided, pending results. Patient-readable summary at end.

Drafts discharge summaries.

💡

Pro tip: Patient-readable summary at the end is what families actually read. Medical-grade body for the chart; plain-English summary for the human. Both matter.

Procedure Note Template

4/20

Procedure note for [procedure name]. Output template: indication, consent obtained, time-out completed, anesthesia, technique step-by-step, findings, complications (none / list), specimens, estimated blood loss, condition at end, post-procedure plan, attending presence. Fill placeholders with [bracketed prompts].

Generates procedure note templates.

💡

Pro tip: Templates with bracketed prompts > free-text. Forces every required element. Audits don't care about your prose; they check elements documented.

XML tags are just the start. Learn the full Claude workflow.

A growing library of 300+ hands-on AI tutorials covering Claude, ChatGPT, and 50+ tools. New tutorials added every week.

Start 7-Day Free Trial

Patient Communication

4 prompts

Patient-Friendly Diagnosis Explanation

5/20

Explain [diagnosis] to a patient at [reading level / education level]. Output: what it is (analogy), why we think they have it, what it means for their life, what we'll do, common questions. Avoid jargon entirely. Use specific everyday words. Tone: serious but not scary.

Explains diagnoses in plain language.

💡

Pro tip: Patient at 6th-grade reading level handles 6th-grade vocabulary about complex conditions. Don't equate vocabulary with intelligence — they're scared, not stupid. Plain words = better understanding + retention.

After-Visit Summary

6/20

Patient-readable after-visit summary. Diagnoses today: [list]. Plan: [list]. Output: what we discussed (plain English), what to do at home (numbered, specific), warning signs (call back if), next appointment, questions to bring next time, prescription instructions in patient terms. Under 1 page.

Writes after-visit summaries.

💡

Pro tip: Patients forget 50% of what you said by the time they hit the parking lot. Written summary = compliance jumps. Make warning signs visually distinct (bold, separate section).

Difficult News Conversation Script

7/20

Help me prepare to deliver [difficult news] to a patient. Output: SPIKES protocol-aligned script — Setting prep, Perception check, Invitation, Knowledge sharing (warning shot then info in chunks), Emotion response, Strategy + summary. Specific words. Where to pause for patient response.

Scripts difficult news conversations.

💡

Pro tip: SPIKES is taught for a reason. The "warning shot" + chunked info pattern lets the patient process. Dumping everything at once = patient hears word 1 and goes deaf for the rest.

Patient Email Response

8/20

Patient sent: [paste message]. Draft a clinically appropriate response. Address their concern, set expectations on what email is for vs what needs a visit, document plan. Tone: warm but boundaried. Avoid promises or definitive remote advice for issues needing exam. Suggest visit if appropriate.

Drafts patient portal responses.

💡

Pro tip: Email = communication, not consultation. Patients sometimes describe emergencies in casual emails. Draft response should triage — does this need a visit, ER, or just info?

Literature + Decision Support

4 prompts

Literature Quick Review

9/20

[Paste abstract or full paper]. Critique this study for clinical practice integration. Output: study design strengths/weaknesses, population (does it match my patients?), effect size in absolute terms (NNT), conflicts of interest, where this fits with current guidelines, would I change practice based on this? Honest take.

Reviews medical literature critically.

💡

Pro tip: Effect size in absolute terms (NNT) > relative risk reduction. RRR misleads. NNT tells you how many patients you'd treat to benefit one. Often unimpressive when stated honestly.

Guideline Reconciliation

10/20

Two guidelines say different things about [clinical question]. [Paste guideline A] vs [paste guideline B]. Output: where they agree, where they diverge + why (different evidence base, different risk tolerance, different population), what to do when they conflict, my judgment factors. Don't just pick one — illuminate.

Reconciles conflicting guidelines.

💡

Pro tip: Different specialties weight evidence differently. AHA vs ACC vs ACP often diverge. Knowing WHY they diverge = clinical judgment. Picking the one that matches your specialty's preference = lazy.

Drug Interaction Brainstorm

11/20

Patient med list: [list]. New rx considering: [drug]. Brainstorm interactions: pharmacokinetic (CYP, renal, hepatic), pharmacodynamic (additive effects, antagonism), specific ones to verify in pharmacy database before prescribing. Patient-specific factors: [age, renal, hepatic]. Always verify against database.

Brainstorms drug interactions.

💡

Pro tip: Brainstorm tool, not source of truth. Use Claude to surface candidates ("what could I be missing?") then verify in Lexicomp/Epocrates. Hallucinated interactions exist; verification non-negotiable.

Risk Stratification Helper

12/20

Risk stratify this patient for [condition/procedure]. Patient: [details]. Output: validated risk tools applicable (mention which), input variables I need to confirm, estimated risk band, what changes management, where the tool's limitations matter for this patient. Frame for clinical decision-making.

Walks through risk stratification.

💡

Pro tip: Risk calculators have validated populations. Asian/elderly/pregnant often poorly represented. Tool's output ≠ this patient's actual risk. Knowing the tool's population matters more than knowing the tool exists.

These prompts give you the what. Tutorials give you the why.

Learn when to use extended thinking, how to build Claude Projects, and workflows that compound. 300+ tutorials and growing.

Try AI Academy Free

Practice + Workflow

4 prompts

Inbox Triage Protocol

13/20

Build an inbox triage protocol for my practice. Volume: [#/day]. Staff support: [describe]. Output: messages MA can resolve, messages requiring nurse review, messages requiring me, escalation criteria, response time targets per category. Specific rules — not vague guidance.

Builds inbox triage protocols.

💡

Pro tip: Untriaged inbox = physician burnout fuel. Specific rules ("MA handles X, escalates if Y") = staff actually act. Vague rules = everything ends up on physician.

Pre-Visit Planning

14/20

Pre-visit prep for tomorrow's schedule: [paste schedule with chief complaints]. For each patient: what to review beforehand, likely workflow (results pending, refills due, screening overdue), efficiency moves (combine visits, hand off to MA, prep orders). 5-10 min total prep — not 30.

Plans efficient pre-visit reviews.

💡

Pro tip: Pre-visit prep saves more time than it takes. 5 minutes the night before = 15 minutes per visit reclaimed. Not prepping = reactive scrambling all day.

Coding + Documentation Audit

15/20

[Paste de-identified note]. Audit for E/M coding: which level does the documentation support, which elements are missing for higher level (if appropriate), what risk-of-complication justification is in the note, is the time documented if billing time-based. Compliance check, not coding-up advice.

Audits notes for coding compliance.

💡

Pro tip: Down-coding from missed documentation is the silent revenue loss. Regular self-audit = catch the patterns. Don't code up beyond what's documented; document up to what you actually did.

Burnout Check-In

16/20

Walk me through a self check-in on burnout. Ask sequentially: emotional exhaustion (1-10), depersonalization (do I feel cynical about patients?), personal accomplishment (do I feel effective?), sleep, exercise, last vacation, last social connection outside work. Reflect what I share — pattern, not just data.

Self-assesses physician burnout risk.

💡

Pro tip: Maslach Burnout Inventory dimensions matter. Exhaustion alone is normal; exhaustion + cynicism + low accomplishment = burnout. Naming the specific dimension = pointed intervention.

Frequently Asked Questions

Standard consumer Claude is NOT HIPAA-compliant. Do not paste PHI. For HIPAA workflows, use Anthropic via AWS Bedrock with BAA, or platforms like Suki/Doximity/Abridge that have BAAs. De-identify before any consumer-Claude use.
No, and shouldn't. Claude assists with documentation, brainstorming differentials, patient communication. Diagnosis = your clinical judgment + exam + history + tests. Claude as cognitive aid is appropriate; as autonomous diagnostician is not.
Yes — drug doses, interactions, study findings can be wrong. Always verify against authoritative sources (Lexicomp, UpToDate, primary literature) before clinical use. Treat Claude as smart-but-unreliable resident — useful, but you sign the order.
For PHI workflows: Anthropic via Bedrock with BAA, or vendor-integrated tools (Suki, Abridge, Nuance DAX) that have BAAs. For non-PHI tasks (literature review, education content, admin templates) — consumer Claude is fine.
No, but it changes the job. Documentation + literature review + admin shrink. Patient relationship + clinical judgment + procedural skill don't. Physicians who use AI well will outperform those who don't. Same as every prior tech wave.

Prompts are the starting line. Tutorials are the finish.

A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.

7-day free trial. Cancel anytime.