Claude Project: Your Personal CDI Intelligence System

What This Builds

A persistent Claude Project that functions as your personal CDI intelligence system — a single place where you can query all your reference materials, draft appeals with cited guidelines, analyze denial patterns, and get answers grounded in the actual documents your program uses. Unlike a one-off conversation, this project remembers its context across every session, so it gets more useful the more documents you add to it.

The result: instead of having AHIMA guidelines in one PDF, your facility policy in another folder, Coding Clinic references in a binder, and ICD-10 guidelines on CMS.gov — you have one conversational interface that knows all of it.

Prerequisites

• Claude {{tool:Claude.plan}} subscription ({{tool:Claude.price}}) — Projects requires paid tier
• Comfortable using Claude for basic document analysis (Level 3)
• Reference documents assembled (see list below)

The Concept

A Claude Project is a persistent workspace where your uploaded documents stay available across every conversation you have within that project. Think of it like a well-organized filing cabinet that can read, and you can ask it questions in plain English. Every time you open a new conversation in the CDI project, Claude already has access to everything you've uploaded — your guidelines, your policies, your denial patterns — without you needing to re-paste anything.

This is different from a Custom GPT in one key way: Claude Projects are better for long-document analysis and nuanced reasoning. If you need to compare how your facility's query policy aligns with AHIMA guidelines, or analyze a complex denial letter against ICD-10 guidelines, Claude Project is more capable for that kind of work. Custom GPT (Level 4 guide) is better for quick, standardized query drafts your whole team uses.

Build It Step by Step

Part 1: Create the Project

1. Go to {{tool:Claude.url}} and sign in
2. In the left sidebar, click Projects → New Project
3. Name it: "CDI Intelligence System"
4. Add a description: "CDI reference system — AHIMA guidelines, ICD-10, facility policy, denial analysis"

Part 2: Write Persistent Project Instructions

Click the Edit project instructions area. This text is always active in every conversation in this project:

You are a CDI Intelligence System for an experienced Clinical Documentation Improvement Specialist.

Your knowledge base contains: AHIMA physician query guidelines, ICD-10-CM Official Guidelines, facility CDI policies, denial analysis data, and specialty-specific coding references.

How to respond:
- Always cite the specific document and section that informs your answer
- When a question isn't covered by uploaded documents, say so explicitly and note you're drawing on training knowledge
- Use clinical CDI terminology — this user is an experienced CDI professional
- For guideline questions: quote the relevant passage directly, then explain its practical application
- For query drafting: follow AHIMA compliance standards from the uploaded guidelines
- For denial analysis: compare the denial rationale against the relevant coding guidelines and identify the strongest counter-arguments
- Format responses for quick scanning: use headers and bullets for multi-part answers

This system handles NO PHI. All case information must be de-identified before being shared here.

Part 3: Build Your Document Library

Upload documents in this priority order:

Tier 1 — Upload first (highest impact):

• AHIMA Physician Query Compliance Guidelines (2019 or current)
• ICD-10-CM Official Guidelines for Coding and Reporting (current fiscal year) — download free from CMS.gov
• Your facility's CDI physician query policy

Tier 2 — Upload next:

• ICD-10-CM Tabular Addenda for the current fiscal year (new/changed codes)
• Your CDI program's top DRG targets and CC/MCC opportunity reference sheet (create a simple Word doc listing your program's 15–20 highest-priority diagnoses and their documentation requirements)
• ACDIS clinical documentation standards reference (if you have it)

Tier 3 — Add as you encounter them:

• Specific Coding Clinic guidance on your most-queried diagnoses (copy relevant passages into a text file)
• Your payer-specific denial patterns summary (de-identified aggregate data only)
• Specialty-specific documentation guides you've created

For each document: click Add content → Upload files → wait for processing.

Part 4: Test with Real Scenarios

After uploading, test with these queries:

Guideline reference test: "What does the AHIMA guideline say about querying for clinical validation denials? Quote the relevant passage."

Cross-document test: "How does our facility query policy align with AHIMA standards for the timing of concurrent queries? Note any gaps."

Practical application test: "Based on the ICD-10 guidelines, what documentation is required to code both acute and chronic respiratory failure as a combination code?"

What you should see: Answers that cite specific document sections and quote relevant passages — not general AI knowledge.

Part 5: Create a Denial Analysis Workflow

Once your documents are loaded, you can analyze denial patterns systematically. Create a running document (Word or Google Doc) that you update monthly:

De-identified denial log template:

Month: [Month/Year]
Denial category: [e.g., sepsis coding, respiratory failure specificity]
Number of denials: [X]
Payer: [Medicare, Medicaid, Commercial - no payer names]
Common denial rationale: [paste the denial language used]
Our appeal outcome: [approved/denied/pending]

Then ask the CDI Project: "Based on this month's denial patterns, which ICD-10 guidelines most directly support our coding positions for the [category] denials? What is our strongest argument for each?"

Part 6: Build a Monthly Update Habit

At the start of each month:

1. Upload any new Coding Clinic guidance for your high-priority diagnoses
2. Add updated ICD-10 tabular changes as they're released
3. Ask: "What's new in this month's uploaded content that affects our query templates?"

Real Example: Analyzing a Contested Sepsis Denial

Setup: Your CDI Project has AHIMA guidelines and ICD-10-CM Official Guidelines uploaded.

Input: "We received a clinical validation denial for sepsis (A41.9). The denial says: 'Documentation does not clearly support a diagnosis of sepsis separate from the underlying infection.' The patient had: UTI as primary infection, SIRS criteria met (temp 38.9, HR 112, WBC 17,500), IV antibiotics within 4 hours. Please identify the strongest arguments from the uploaded guidelines to support our appeal."

Output: A structured analysis that quotes: (1) the ICD-10-CM Official Guidelines Section I.C.1.d on sepsis coding requirements, (2) the AHIMA guidance on clinical validation query standards, and (3) suggests specific language for the appeal letter based on the documented clinical criteria — all cited to the specific documents you uploaded.

Time saved: 45 minutes of manual reference searching → 3 minutes of conversation.

What to Do When It Breaks

• Claude can't find the answer in uploaded documents → It will tell you explicitly, which is correct behavior; verify the document was uploaded successfully by checking the project knowledge list
• Large PDF fails to upload → Split into sections or copy/paste key passages as a text file
• Answers are too generic → Add more specific documents; the system is only as good as what you've given it
• Context gets confused across a long conversation → Start a new conversation within the project; your documents remain available

Variations

• Simpler version: Skip Tier 3 documents and use only AHIMA guidelines + ICD-10 Official Guidelines — covers 80% of use cases with 30 minutes of setup
• Extended version: Add your payer contracts' coding language requirements and build a payer-specific denial response workflow

What to Do Next

• This week: Upload Tier 1 documents and test 5 reference questions you commonly need to look up manually
• This month: Add Tier 2 documents and run your first denial pattern analysis
• Advanced: Connect your CDI Project insights to your team's education calendar — use the system to identify which physician documentation gaps are generating the most denials, then target education at those gaps

Advanced guide for Clinical Documentation Improvement Specialist professionals. Requires Claude {{tool:Claude.plan}} subscription ({{tool:Claude.price}}).