How to Use Copilot Prompts for Healthcare Patient Summaries

Healthcare professionals often need to generate concise patient summaries from clinical notes, lab results, and visit records. Copilot can help transform scattered medical data into a structured summary, but only if you write the right prompt. Many users find that vague prompts produce incomplete or irrelevant output. This article explains how to structure prompts for patient summaries, what data sources Copilot reads, and how to avoid common pitfalls.

You will learn the exact phrasing and formatting rules that produce accurate, HIPAA-aware summaries. The guide covers prompt templates for admission notes, discharge summaries, and follow-up visits. You will also see how to restrict Copilot to specific Microsoft Graph data to prevent hallucinated or outdated information.

How Copilot Handles Medical Data for Summaries

Copilot in Microsoft 365 uses the Microsoft Graph API to access data from Exchange, SharePoint, OneDrive, and Teams when you are signed in with a work or school account. For healthcare summaries, Copilot can read patient notes stored in Word documents, lab results in Excel tables, and visit logs in SharePoint lists. It cannot read data from external electronic health record systems unless those systems are connected via Microsoft Cloud for Healthcare or a custom Graph connector.

The quality of the summary depends on three factors. First, the prompt must define the role and audience. Second, the prompt must specify which data to include and which to exclude. Third, the prompt must request a structured format such as bullet points, sections, or a table. Without these elements, Copilot may generate a narrative that mixes subjective observations with objective findings or includes data from the wrong patient record.

Copilot does not perform clinical reasoning or diagnosis. It extracts and rearranges text from the documents you provide. You must verify every summary for accuracy before adding it to a patient chart or sharing it with a colleague.

Steps to Write a Prompt for a Patient Summary

Follow these steps to create a prompt that produces a structured, accurate patient summary. The example uses a fictional patient named Jane Doe with a recent hospital admission for pneumonia.

1. Open the source documents in Copilot
   Open the Word document or SharePoint page that contains the patient notes, lab results, and medication list. Copilot can read the content of the file you have open when you use the “Summarize this file” command. For multiple files, open each one in a separate tab and reference them by name in the prompt.
2. Set the role and audience in the first sentence
   Type: “You are a clinical documentation specialist. Write a summary of the patient records below for a primary care physician.” This tells Copilot to use professional medical language and to focus on information relevant to a physician, such as diagnosis, treatment, and follow-up needs.
3. Define the data scope explicitly
   Add: “Use only the information from the file ‘Jane Doe Admission Notes.docx’ and ‘Jane Doe Lab Results.xlsx’. Do not include information from any other files or from general knowledge.” This prevents Copilot from pulling data from unrelated documents in your OneDrive or SharePoint.
4. Request a structured format
   Add: “Organize the summary into these sections: Chief Complaint, History of Present Illness, Relevant Lab Results, Medications Administered, and Discharge Plan. Use bullet points under each section.” This ensures the output is scannable and easy to paste into a clinical note template.
5. Set a length limit
   Add: “Limit the summary to 250 words.” Without a limit, Copilot may generate a long narrative that repeats information. A word limit forces it to prioritize the most important data.
6. Review and edit the output
   Read the generated summary. Verify that every lab value matches the source file. Correct any misinterpreted abbreviations or missing data. Copy the final text into your electronic health record system or clinical note template.

Prompt Template for Discharge Summary

Use this template when you need a discharge summary from a hospital stay. Replace the bracketed placeholders with your patient's details.

“You are a clinical documentation specialist. Write a discharge summary for [Patient Name] based only on the file ‘[File Name.docx]’. Include these sections: Admission Date, Discharge Date, Primary Diagnosis, Procedures Performed, Medications at Discharge, and Follow-Up Instructions. Use full sentences for the Diagnosis and Procedures sections. Use bullet points for Medications and Follow-Up. Limit the summary to 300 words.”

Prompt Template for Follow-Up Visit Note

Use this template for a brief outpatient follow-up summary. It produces a SOAP-style note.

“You are a medical scribe. Write a SOAP note for [Patient Name] using only the data from ‘[File Name.docx]’. Include Subjective, Objective, Assessment, and Plan sections. Under Objective, include vital signs and relevant lab results. Under Plan, list medications and next appointment date. Use bullet points under each section. Limit the summary to 200 words.”

If Copilot Returns Incomplete or Incorrect Summaries

Even with a well-written prompt, Copilot may produce summaries that miss key details or include incorrect information. The following issues are common in healthcare settings.

Copilot omits lab values or vital signs

This usually happens when the lab data is in a table format that Copilot does not parse correctly. To fix this, open the Excel file or Word table and select the cells you want to include. Then use the command “Summarize the selected content” instead of referencing the whole file. Alternatively, paste the lab values as plain text into a new document and reference that file.

Copilot includes data from the wrong patient

If you have multiple patient files open or in your recent documents, Copilot may mix data. Always close all unrelated files before starting a prompt. Use the phrase “based only on the file [exact file name]” in the prompt. If the problem persists, remove the file from your recent list by clicking File > Open > Clear unpinned files.

Copilot uses outdated medication lists

Copilot reads the current version of the file. If the medication list was updated in a different version or a separate document, Copilot will not see it. Ensure that the source file you reference contains the most recent medication list. If the list is in a separate file, include both file names in the prompt and specify which sections to pull from each file.

Copilot Prompt for Summary vs Manual Dictation: Key Differences

Item	Copilot Prompt for Summary	Manual Dictation
Time to produce a 250-word summary	30 to 60 seconds	5 to 10 minutes
Requires structured prompt	Yes, must define role, scope, and format	No, the clinician organizes thoughts verbally
Risk of data mixing	Moderate if multiple files are open	None, since the clinician controls the source
Ability to pull lab values from tables	Good if cells are selected or table is simple	Requires manual lookup
Output consistency	High when the same prompt template is reused	Varies by clinician and fatigue

Copilot reduces the time needed to draft a summary, but it requires upfront effort to write and test prompts. Manual dictation gives full control but takes longer and depends on the clinician's memory. Most healthcare teams use a hybrid approach: Copilot generates a first draft, then the clinician edits and verifies the content.

After you have tested the prompt templates in this article, try creating a custom prompt that includes your facility's preferred headings or abbreviations. Save the working prompt in a OneNote page or a SharePoint document so you can copy it for each patient. Always verify the output against the original source files before using the summary in a clinical setting. The most reliable method is to keep the source document open side by side with the Copilot pane and check each section as you read it.