Prompt Chain: Prior Authorization End-to-End Workflow

What it accomplishes for Medical Assistants

Tools: ChatGPT Plus or Claude Pro | Time to build: 1-2 hours | Difficulty: Intermediate Prerequisites: Comfortable using ChatGPT Plus for PA letters — see Level 3 guide: "Set Up ChatGPT Plus for Prior Authorization Letters"

What This Builds

Instead of handling each prior authorization denial as a separate, from-scratch task, this prompt chain walks you through a structured 4-step process: analyze the denial → identify the strongest argument → draft the appeal → format it for sending. By chaining these steps together, each output feeds the next — and you consistently produce stronger appeals with less effort.

Prerequisites

• ChatGPT Plus or Claude Pro ($20/month) — both work equally well for this
• Completed the Level 3 guide on PA letter setup (Custom Instructions or Claude Project)
• A real denied prior authorization to practice with
• Cost: $20/month (same subscription you're already using)

The Concept

A prompt chain is like a production line. Instead of asking one question and getting one answer, you ask a series of connected questions where each answer becomes an input for the next. For a prior authorization appeal, it works like this:

Step 1 → AI analyzes the denial reason Step 2 → AI identifies the strongest counter-arguments Step 3 → AI drafts the appeal using those arguments Step 4 → AI formats the final letter

Each step is focused and builds on the previous one — producing a much stronger result than a single "write me an appeal letter" prompt.

Build It Step by Step

Part 1: Set Up the Chain

You'll run 4 consecutive prompts in a single conversation. Open ChatGPT Plus or Claude Pro and keep the window open throughout — the AI remembers your previous messages.

Save this chain template in a Google Doc or Notes app for reuse:

PROMPT CHAIN: Prior Authorization Appeal

STEP 1 — ANALYZE:
I have a prior authorization denial. Here are the details:
- Procedure/Medication: [fill in]
- Diagnosis (ICD-10): [fill in]
- Insurance company: [fill in]
- Denial reason (exact language from denial letter): [fill in]
- Clinical history I have documented: [fill in what's in the chart]

Analyze this denial. What is the insurer's likely concern? What documentation gaps might have caused this denial? What does the insurer typically want to see for this type of request?

STEP 2 — STRATEGY (paste after Step 1 response):
Based on your analysis, what are the 3-4 strongest arguments for medical necessity in this case? What clinical evidence, guidelines, or documentation should I include in the appeal? Are there any CMS or clinical practice guidelines that support this request?

STEP 3 — DRAFT (paste after Step 2 response):
Now draft a formal prior authorization appeal letter using those arguments. Include: formal letter format, medical necessity language, the clinical arguments you identified, reference to the patient's treatment history, and a specific request for reconsideration. One page maximum. Include placeholders for: [PHYSICIAN NAME], [NPI], [PATIENT NAME], [DATE OF BIRTH], [CLAIM/AUTH NUMBER].

STEP 4 — REFINE (paste after Step 3 response):
Review the draft letter. Make these improvements: strengthen the medical necessity argument, ensure the denial reason is directly addressed in the first paragraph, add a closing line requesting expedited review if clinically appropriate, and confirm all required elements are present (patient info placeholders, physician signature line, appeal date). Provide the final clean version.

Part 2: Run the Chain with a Real Case

1. Open ChatGPT Plus or Claude Pro → New conversation
2. Paste Prompt 1 (ANALYZE) with your actual case details filled in
3. Read the response carefully — note what the AI identifies as the key issues
4. Paste Prompt 2 (STRATEGY) immediately after in the same conversation
5. Read the strategy response — make note of any guidelines or evidence you can verify
6. Paste Prompt 3 (DRAFT) — get your draft letter
7. Paste Prompt 4 (REFINE) — get the final polished version

Part 3: Verify and Finalize

After Step 4, you have a polished appeal letter. Before sending:

1. Fill in all [PLACEHOLDERS] with actual patient information
2. Verify any clinical guideline citations the AI mentioned (search the guideline name to confirm it exists and supports your case)
3. Have your physician review — they sign off on the clinical rationale
4. Add your practice letterhead
5. Submit via fax, payer portal, or mail per the payer's instructions

Real Example: Cardiology Referral Denial

The situation: A patient with unexplained palpitations and a borderline EKG was referred to cardiology. The insurer denied the referral, citing "primary care management should be attempted first."

Step 1 prompt (filled in):

I have a prior authorization denial. Here are the details:
- Procedure: Cardiology consultation (CPT 99243)
- Diagnosis: R00.2 — Palpitations; R94.31 — Abnormal EKG
- Insurance: BlueCross BlueShield
- Denial reason: "Condition should be managed at the primary care level before specialist referral"
- Clinical history: Patient has had recurrent palpitations for 3 months. EKG shows borderline QT prolongation. Symptoms affecting daily activity. PCP has reviewed and determined specialist evaluation needed given abnormal EKG finding.

Analyze this denial. What is the insurer's likely concern? What would strengthen this appeal?

Step 1 output (example): AI identifies that BCBS typically wants to see: documented frequency of symptoms, any medications already tried or contraindicated, and clinical rationale for why the EKG finding warrants specialist review rather than watchful waiting.

Step 2 prompt:

Based on your analysis, what are the 3-4 strongest arguments for medical necessity in this case? What clinical evidence or guidelines support a cardiology referral for documented QT prolongation with symptomatic palpitations?

Step 2 output: AI provides: (1) ACC/AHA guidelines on evaluation of palpitations with EKG abnormalities, (2) QT prolongation risk stratification rationale, (3) primary care limitations in interpreting complex EKG findings, (4) risk of adverse cardiac event without proper evaluation.

Step 3 + 4: Draft and refine produces a letter that opens with the specific denial reason addressed directly, cites ACC/AHA guidelines, explains the EKG finding in clinical terms, and closes with an urgent request for reconsideration given the cardiac risk.

Time: 4 prompts, 15 minutes total. Better result than 45 minutes of drafting from scratch.

What to Do When It Breaks

• "The AI doesn't know the specific payer's requirements" → Add context: "This is a UnitedHealthcare Community Plan appeal. They typically require..." If you don't know their specific requirements, the AI can suggest common payer requirements, but you should verify against the actual denial letter and payer website.

• "The AI cited a guideline I can't verify" → Ask: "Can you give me the exact name and section of the clinical guideline you cited?" Then search for it. If it doesn't exist, ask the AI to find a real alternative.

• "Step 3 draft is too long" → Add to your Step 4 prompt: "Also shorten the total letter to strictly one page — cut any redundant language."

• "Physician doesn't like the tone" → Add to Step 4: "Revise to be [more aggressive/more collegial/more data-focused]. The physician prefers [description of preferred style]."

• "The same insurer keeps denying the same procedure" → Build a "denial pattern document" — each time you win an appeal, paste a de-identified copy of what worked into a Google Doc. Use that as context in Step 1 for future denials from the same payer.

Variations

• Simpler version: If the denial is straightforward, skip Step 1 and 2 and go directly to Step 3 with your draft prompt. Use Step 4 only to clean up the language.

• Extended version: Add a Step 5 for peer-to-peer prep: "I have a peer-to-peer call scheduled with the medical director. Based on the letter we just drafted, give me 5 concise talking points and likely pushback questions with suggested responses."

What to Do Next

• This week: Run your first real denial through this chain; save the final letter to your "successful appeals" library
• This month: Track which payers respond best to which argument types; build a payer-specific note in your Step 1 prompt
• Advanced: Create a Google Doc "PA Denial Library" — when you win an appeal, de-identify it and add it. Use it as additional context in Step 1 for similar future denials.

Advanced guide for medical assistant professionals. All appeal letters must be reviewed and signed by the supervising physician before submission. AI provides drafting support; clinical judgment is the physician's responsibility.