A RAC Letter Shouldn't Be the First Time You See a Problem Chart.
Every Medicare chart audited against CMS rules and reviewed for documentation sufficiency before claims go out — not after a recoupment demand lands on your desk.
What's at Stake
Medicare audit exposure compounds quickly across a year of charts — and most of it never gets a second look before claims go out.
- Per-violation penalty range — stacked across a year of charts
- $100–$50,000
- Share of charts a typical coder reviews manually before billing
- ~10%
- Of RAC-reviewed claims found to contain improper payments
- ~20%
Per-violation penalty range — stacked across a year of charts
Share of charts a typical coder reviews manually before billing
Of RAC-reviewed claims found to contain improper payments
CAAS Audits 100% of Your Medicare Charts Before They're Billed.
CAAS — the Chart Audit Automation System — is software that reads every chart you're about to submit, checks it against a 9-rule engine built on CMS LCD/NCD guidance, and uses AI to judge whether your documentation actually supports the code billed. Charts get a green/yellow/orange/red risk score, so your coders spend their time on the 15 charts that need attention instead of skimming all 200. Cross-chart pattern detection catches the things no human reviewer can spot at scale — clone notes, upcoding drift, time overlaps, stale templates. The black box becomes a triage queue.
From Chart Batch to Clean Claims in Four Steps
No workflow rebuild. Upload after your providers close their notes, review the flags, then submit.
- 01
Upload
Bulk-upload a chart batch as CCDA files — BAA must be on file before ingest is enabled.
- 02
Audit
The 9-rule engine and AI documentation review evaluate each chart section by section.
- 03
Risk-Score
Every chart gets a green, yellow, orange, or red rating with the specific reasons each flag fired.
- 04
Review
Your coders triage only the flagged charts — confirm, dismiss, or escalate before billing.
Six Layers of Compliance Check on Every Chart
Each layer catches a different failure mode — from missing signatures to coder-blind cross-chart patterns.
Smart Chart Ingestion
Bulk-upload chart batches in CCDA format — the system parses problem lists, HPI, exam, assessment, and plan into structured fields the audit engine can actually evaluate. BAA-gated at the database level.
9-Rule Compliance Engine
Encodes CMS LCD/NCD requirements, provider roster and NPI scope-of-practice checks, CPT-ICD pairing logic, signature timeliness, and medical necessity tests — the rules a MAC reviewer would apply, applied first.
AI Documentation Review
Anthropic Claude reads each note section and rates whether the documentation supports the billed service — flagging insufficient HPI, thin exam, or a plan that doesn't justify the complexity level.
Color-Coded Risk Scoring
Green, yellow, orange, red. Coders see at a glance which charts are clean and which need a closer look — instead of treating every chart as either a sample or a black box.
Cross-Chart Risk Analytics
Detects patterns no individual chart review can catch — clone notes across patients, upcoding drift against your own clinic baseline, daily volume anomalies, and provider time overlaps.
Reviewer Dashboard & Workflow
Flagged charts route into a triage queue where coders mark each one confirmed, dismissed, or escalated — every action timestamped to the 29-event HIPAA audit trail.
The Patterns RAC Auditors Look For — And Your Coders Can't See
RAC and TPE reviewers analyze your billing across hundreds of charts at once. They're not reading one note in isolation — they're looking for patterns. Neither should your audit process.
Clone-Note Detection
Flags HPI sections that are more than 85% identical across different patients. A coder reviewing one chart at a time has no way to know the same paragraph appeared in four other patients last week — but the auditor pulling a 50-chart sample will see it instantly.
Upcoding Detection
Compares each provider's high-complexity code rate against the clinic baseline and flags providers running more than 2× the average. This is exactly the pattern that triggers a TPE probe. Catching it internally — before claims go out — is the entire point.
Volume Anomaly Detection
Flags providers exceeding plausible per-day encounter thresholds. A psychiatrist billing 35 90834s in a day, or a primary-care provider with 28 99214s, will get a MAC's attention. The system flags it before submission, not after recoupment.
Doc-Length vs. Code-Complexity Check
A 50-word progress note billed as 99310 (high-complexity nursing facility care) is a documentation deficit no defense can save. CAAS checks each note's substance against the complexity level claimed — and rejects the mismatch.
Stale Template Detection
Flags three or more consecutive notes for the same patient that are functionally identical. This is the cloned-template problem — providers copying yesterday's note forward without updating the clinical picture. RAC auditors call it out immediately.
Time-Overlap Detection
Flags any provider billing two patient encounters that overlap in time. Impossible-on-its-face billing is one of the cleanest audit findings there is — and one of the easiest to prevent if you actually look for it.
Your Patients' Data Stays Safe
-
Charts Encrypted End to End
Every chart is scrambled in transit and locked down in storage — unreadable to anyone outside your clinic.
-
Full Activity Log
Every view, edit, and decision is logged with a permanent record nobody can quietly alter.
-
HIPAA-Compliant by Design
Built around HIPAA rules from day one — not bolted on later. Designed to pass the same audits your clinic has to pass.
-
Your Charts Aren't AI Training Data
The AI that reviews each chart never trains on it. Your patient notes don't become part of anyone else's model.
Built for Any Clinic That Bills Medicare
The compliance rules don't change based on specialty — and neither does CAAS.
FQHCs and Look-Alikes
Primary care and behavioral health under one roof means twice the code families, twice the LCD scrutiny — G-codes, 99213/99214, 90791, all in one billing run.
Behavioral Health Groups
90791, 90792, 90832, 90834 — high-volume codes with strict documentation requirements that auditors target precisely because the notes tend to look alike.
Psychiatric Practices
Med-management E/M plus psychotherapy add-on codes are a documented MAC focus area — the exact CPT-ICD pairing patterns CAAS rules check on every chart.
Nursing Facility Care Providers
99308, 99309, 99310 — complexity levels that have to be supported by the note itself, not by the patient's overall acuity. CAAS checks the substance, not the assumption.
Frequently Asked
Which EHR systems does CAAS support?+
CAAS ingests charts as CCDA files, which most major EHRs can export. Adding native support for a specific EHR's proprietary format is a custom-development task scoped during implementation. Bring us your EHR and chart-volume profile and we'll tell you exactly what the integration looks like.
How does CAAS handle our BAA?+
A signed BAA is required before any chart can be uploaded — the gate is enforced at the database level, not the UI. Our standard BAA covers the AI documentation-review subprocessor (Anthropic) and all storage and access controls. Custom BAA language is reviewable on request.
What's the implementation timeline?+
Plan on 4–8 weeks for a typical install. Timeline depends on your EHR's CCDA export capability, BAA execution, user-role mapping, and how much rule customization your specialty requires. We scope all of this on the first call.
Can it audit non-Medicare claims?+
The 9-rule engine is anchored in CMS LCD/NCD guidance, so the strongest fit is Medicare and Medicare Advantage. Many of the documentation-sufficiency and cross-chart pattern checks apply equally to commercial and Medicaid claims, but coverage rules differ and we'll be upfront about which checks transfer.
Who reviews the flagged charts — your team or ours?+
Yours. CAAS surfaces what needs attention and explains why each flag fired; your coders and compliance team triage. The product is designed to make your existing reviewers faster and more thorough, not to replace clinical judgment with a vendor.
How does pricing work?+
Pricing is custom per clinic, based primarily on chart volume and the rule customization required for your specialty. There's no per-seat tier and no published price list — talk to us about your volume and we'll quote it directly.
Is the AI's evaluation auditable and explainable?+
Yes. Every AI review includes section-by-section ratings (history, exam, medical decision-making, plan), specific feedback on what's missing, and a confidence score. Admins can see the prompt templates the AI was given. A flag is never a black box — you can show an auditor exactly why each chart was rated the way it was.
Stop Finding Out About Problem Charts From CMS
A 30-minute demo, walked through with your chart types and codes. We'll show you what 100% audit coverage looks like on real Medicare workflows — and what it would catch in your clinic.