Writing a Discharge Summary with AI

5 min read

The Problem

Discharge summaries take 15–20 minutes per patient on average — often more for complex admissions. That’s time taken from patient care, teaching, or rest. AI can reliably cut this to under 5 minutes while improving consistency and completeness.

This workflow walks you through a repeatable 3-step process for drafting discharge summaries using Claude or a similar LLM.


Step 1: Gather Your Inputs

Before prompting the AI, collect the key clinical information you’ll need:

  • Admission diagnosis and reason for visit
  • Hospital course (brief narrative of what happened)
  • Key studies (labs, imaging, procedures)
  • Medications changed (started, stopped, dose-adjusted)
  • Discharge condition (stable, improved, etc.)
  • Follow-up instructions (appointments, pending results, return precautions)

Having these organized before you prompt will dramatically improve output quality. A messy input yields a messy output.


Step 2: Use the Prompt Template

Use the template below with Claude. Paste in your de-identified clinical notes or a structured summary of the above points.

Discharge Summary Prompt Claude

You are an experienced hospitalist physician writing a discharge summary. Using the clinical information below, draft a complete, concise discharge summary suitable for the outpatient primary care provider.

Format the summary with the following sections:

  1. Admission Diagnosis
  2. Hospital Course
  3. Significant Results
  4. Discharge Medications (changes only — started, stopped, dose-changed)
  5. Discharge Condition
  6. Follow-up Instructions
  7. Return Precautions

Be concise and clinically precise. Avoid filler phrases. Use plain language for the follow-up section.

CLINICAL INFORMATION: [Paste your de-identified clinical notes here]


Step 3: Review and Edit

The AI output is a first draft, not a final product. Always:

  1. Verify medication accuracy — this is the highest-risk section
  2. Check that pending results are documented and follow-up is assigned
  3. Ensure the hospital course narrative is complete and accurate
  4. Remove any fabricated information (hallucinations are rare but possible)
  5. Sign and co-sign per your institution’s requirements

Claude for Clinical Practice

BAA Available
5.0

Best-in-class for long clinical narratives and nuanced reasoning

Related Content