Custom GPT: Build a Nursing Documentation Assistant for Your Unit
What This Builds
A Custom GPT configured specifically for your unit's patient population, charting standards, and documentation tasks — a persistent AI assistant that already knows your specialty, your most common diagnoses, and how your unit documents. Instead of re-explaining your context every time, you open your Custom GPT and it's ready to help. Nurses who build this report 20–40 minutes of documentation time saved per shift, with better consistency across their charting.
Prerequisites
- ChatGPT Plus subscription ($20/month at chat.openai.com)
- Comfortable using ChatGPT for SBAR notes, patient education, and care plans (Level 3)
- 1–2 hours for initial build and testing
- Your unit's common diagnoses, typical patient types, and documentation standards in mind
The Concept
A Custom GPT is like a specialized version of ChatGPT that you configure for a specific purpose. You write a "system prompt" — a set of instructions that runs before every conversation — that tells ChatGPT who it's talking to, what kind of work it helps with, how to format its responses, and what it knows about your context.
Think of it as hiring a new documentation assistant who already knows: your unit type, your 10 most common patient diagnoses, your charting format (SBAR, SOAP, NANDA), and your professional communication standards. Every conversation starts from that shared understanding — you just describe the patient situation and ask for the note.
Build It Step by Step
Part 1: Plan your Custom GPT
Before touching ChatGPT, write down the answers to these questions on paper or your phone:
What unit are you on? (Med-surg, ICU, PCU, ED, L&D, oncology, etc.)
What are your 8–10 most common patient diagnoses? (e.g., CHF exacerbations, post-op hip replacement, COPD exacerbation, sepsis, pneumonia, DVT/PE, diabetic ketoacidosis)
What documentation formats does your unit use? (SBAR for physician calls, SOAP for nursing notes, NANDA for care plans, I-PASS for handoffs?)
What are your top 5 documentation tasks? (Shift handoff notes, patient education materials, care plan goals, incident reports, SBAR call scripts)
What tone do you want? (Professional/clinical vs. conversational; concise vs. thorough)
What should it never do? (Include patient identifiers, give medication doses, make clinical decisions)
Part 2: Create your Custom GPT
- Log into chatgpt.com with your Plus account
- Click "Explore GPTs" in the left sidebar
- Click "Create" in the top right of the GPT store page
- You'll see a "Create" tab and a "Configure" tab. Click "Configure."
Fill in the Configure fields:
Name: "Med-Surg Documentation Assistant" (or your unit type)
Description: "Nursing documentation helper for [your unit] RNs — SBAR notes, care plans, patient education, incident reports"
Instructions (this is the most important part — use this template, customized with your details):
You are a nursing documentation assistant for registered nurses working on a [28-bed medical-surgical / 16-bed PCU / 10-bed ICU] unit at a community hospital.
UNIT CONTEXT:
Common patient types: [post-surgical, CHF, COPD, pneumonia, sepsis, DVT/PE, diabetic management, etc.]
EHR system: Epic
Documentation formats used: SBAR (physician calls and handoffs), SOAP (nursing notes), NANDA (care plans), I-PASS (shift handoffs)
YOUR JOB:
Help draft: shift handoff notes (SBAR/I-PASS format), care plan goals and interventions (NANDA format), patient education materials (5th–6th grade reading level), incident report narratives, SBAR call scripts for physician notification, and professional emails/communications.
FORMATTING RULES:
- Keep documentation concise — nurses are busy
- Use bullet points rather than long paragraphs
- SBAR notes should be under 200 words total
- Care plan goals must be SMART (Specific, Measurable, Achievable, Relevant, Time-bound)
- Patient education must avoid medical jargon
CRITICAL RULES:
- NEVER include real patient names, MRNs, dates of birth, or other identifying information — nurses describe patients in de-identified terms
- NEVER make clinical decisions or medication recommendations beyond what is asked
- ALWAYS add a note when clinical content should be verified (e.g., "verify current dosing with pharmacy")
- If asked something outside nursing documentation, redirect politely
TONE: Professional and clinical. Concise. Practical. Written as if for an experienced nurse.
Leave "Conversation starters" empty or add 3–4 examples:
- "Write a shift handoff SBAR for my post-op patient"
- "Generate care plan goals for fluid volume excess"
- "Help me prepare an SBAR to call the doctor"
- "Write patient education about new warfarin"
Click "Save" → select "Only me" for privacy (don't share your unit-specific GPT publicly)
Part 3: Test and Refine
Test 1: Shift handoff note Open your Custom GPT and type: "Write a shift handoff SBAR. Post-op day 2 after total knee replacement, vitals stable, pain 5/10 on scheduled oral oxycodone, physical therapy ambulated 100 feet this morning, incision sites intact with minimal drainage, awaiting discharge planning consult."
What to check: Is the format SBAR? Is it under 200 words? Is the language clinical and concise? Does it include all the facts you gave it?
Test 2: Care plan Type: "Care plan goals for the nursing diagnosis: Impaired Physical Mobility related to post-surgical pain and muscle weakness as evidenced by limited range of motion and inability to ambulate without assistance."
What to check: Are the goals SMART? Do they have specific timeframes ("patient will ambulate 100 feet by discharge")? Do the interventions match the etiology?
Test 3: Patient education Type: "Patient education handout on new warfarin for a 70-year-old man going home after DVT. 5th grade reading level."
What to check: Is it simple and clear? Does it cover key safety information (INR monitoring, bleeding precautions, diet considerations)? Is it appropriate for print?
Common adjustments:
- If outputs are too long: add "Keep all outputs under [X] words" to your instructions
- If tone is too casual: add "Use formal clinical documentation language"
- If it keeps asking clarifying questions: add "Make reasonable assumptions about common clinical scenarios and provide a response — the nurse will edit"
Real Example: Using Your Custom GPT on a Busy Shift
Setup: You've built a "Med-Surg Documentation Assistant" Custom GPT with your unit context.
Input: It's 6:45pm, 15 minutes before shift change. Patient 1: 58M, post-op day 1 laparoscopic cholecystectomy, stable vitals, pain 3/10 on oral meds, tolerating liquids, ambulated twice, ready for potential discharge tomorrow.
What you type: "Shift handoff SBAR for: 58-year-old male, POD 1 lap chole. Vitals stable. Pain 3/10 oral meds. Tolerating clear liquids. Ambulated twice with encouragement. Incision sites x3, dry and intact. Expected discharge tomorrow pending surgeon approval. Family aware."
What you get: A formatted 120-word SBAR covering all 4 sections, ready to review, edit, and use in your verbal handoff in 30 seconds.
Time saved: 10–12 minutes per patient × 5 patients = 50–60 minutes/shift, once you've practiced the workflow.
What to Do When It Breaks
"The output is too generic and doesn't feel like my unit" → Go back to Configure → Instructions and add more specifics about your patient population and charting expectations. More context = better output.
"It's including clinical advice I didn't ask for" → Add to instructions: "Only provide what is explicitly asked. Do not add unsolicited clinical recommendations."
"It keeps asking me clarifying questions instead of just drafting" → Add: "Proceed with reasonable assumptions and generate a draft. The nurse will review and edit."
"The outputs are too long" → Add explicit word count limits to your instructions for each document type.
"It generated something clinically inaccurate" → This happens. The Custom GPT is a documentation assistant, not a clinical reference. Always verify clinical content — especially medication information — before signing notes.
Variations
Simpler version: Skip the Custom GPT setup entirely and just use a saved text prompt on your phone that you paste into regular ChatGPT each time. Slightly less convenient but functionally similar for low-volume use.
Extended version: Build separate Custom GPTs for different contexts — a "Patient Education Writer GPT" configured specifically for discharge education, a "Certification Study GPT" for exam prep. Each context gets its own optimized setup.
What to Do Next
- This week: Build your first version, test with 10 real documentation scenarios, and refine the instructions based on what you find
- This month: Share the concept with your nurse manager — unit-level Custom GPTs shared among the team create consistency in handoff quality and care plan documentation
- Advanced: Add unit-specific protocol summaries to your GPT's knowledge base using the file upload feature in the Custom GPT configuration
Advanced guide for registered nurse professionals. ChatGPT Plus ($20/month) is required for Custom GPT creation. These tools use more sophisticated AI features and require active review of all outputs before clinical use.