Workflow: Drafting a Patient Education Letter

This workflow takes you from a blank page to a finished patient education letter. It assumes you are using a general-purpose AI tool such as Claude or ChatGPT. The steps are the same for either.

The whole process takes around ten to fifteen minutes once you are familiar with it.

1. Decide what you need before you open the tool

   Before you type anything, spend one minute being clear about the output you want. Ask yourself: Who is this letter for? What does she need to understand? What does she not need to know right now? What tone is right for her? You do not need to write this down — but having it clear in your mind makes the next step much faster.

2. Open your chosen AI tool

   Go to claude.ai or chat.openai.com and sign in. Both have a free tier that is sufficient for drafting documents. You will see a text box. This is where you type your instruction to the tool. It is called a prompt.

3. Write your prompt with four elements

   A good prompt for a patient letter includes four things: the audience (who the letter is for), the purpose (what it needs to explain), the format (length, tone, structure), and any constraints (what to leave out). For example: "Write a one-page letter for a woman in her late forties who has been told she needs a hysteroscopy. Explain what the procedure is, how long it takes, and what she can expect on the day. Use plain English. Do not mention specific sedation or anaesthetic options."

4. Read the first draft critically as a clinician

   The tool will produce a draft in under thirty seconds. Read it as you would read any clinical document before it goes to a patient. Look for: anything clinically inaccurate, anything that could cause unnecessary alarm, anything missing that a patient in this situation genuinely needs to know, and any sentences that are too long or too technical.

5. Refine with a follow-up instruction

   If the draft is not right, do not start again. Tell the tool what to change. For example: "The third paragraph is too technical. Rewrite it in plainer language." Or: "Add a section at the end about what to do if symptoms worsen after the procedure." The tool will rewrite only the part you specify. You can do this as many times as you need.

6. Check tone for your specific patient

   Read the letter again, this time imagining the specific patient — or a typical patient in this situation. Is the tone right for someone who is anxious? For someone who is pragmatic and wants facts? For a patient whose first language may not be English? Adjust the prompt if needed: "Rewrite in simpler language, assuming the reader may have limited health literacy."

7. Add the information only you can add

   The tool does not know your clinic's phone number, your name, your unit's specific follow-up pathway, or your patient's personal circumstances. Copy the draft into your word processor or clinic system. Add these details yourself. This is also the point at which you add the patient's name and details — never in the AI tool.

8. Do a final read before sending

   Read the completed letter aloud — or at least mouth the words. This catches sentences that read well silently but are hard to follow when spoken. A patient reads a letter once, quickly, probably when stressed. Every sentence needs to be immediately clear.

What this means for you

The first time you do this, it will feel slower than just writing the letter yourself. By the third or fourth time, you will be faster. The real gain is not speed — it is consistency. Every patient you see with the same condition gets a letter of the same quality, not one that varies based on how tired you were when you wrote it.

When not to use this workflow

Do not use this workflow for letters that require clinical precision beyond general education — for example, letters explaining specific treatment decisions or communicating results that carry significant clinical implications. In those cases, the letter is a clinical document, and the standards for accuracy are higher than a drafting tool can reliably meet without close review.

Remember: AI is a helpful assistant, not a clinician. You make the call.