How Audra reviews a medical bill.

Every uploaded bill is OCR-extracted into line items (CPT, HCPCS, DRG codes; provider details; charges; adjustments). Each line is then evaluated against three rule sets:

1. Federal billing rules — the CMS NCCI procedure-to-procedure edits, MUE (Medically Unlikely Edits), the No Surprises Act balance-billing protections, and EMTALA-related charges for emergency care. We pull these from the public CMS rule files and the statute text at CMS.gov and CMS No Surprises Act portal.
2. Insurer contracted rates — your plan's in-network allowed amounts and your EOB. We compare provider charges to insurer-allowed amounts to catch overbilling against contract.
3. State consumer-protection law — every state layers its own surprise-billing and balance-billing limits on top of federal protections. We maintain a per-state statute database (visible at /state) reviewed quarterly against the official text.

Every flagged finding cites the specific rule, regulation, or contracted rate behind it. You can verify each one yourself against the linked source — that's deliberate.

The OCR + line-item matching pipeline is fully automated (built on Anthropic Claude + AWS Textract). The AI's job is to read the bill, identify line items, look up the applicable rule, and produce a candidate finding with a severity score.

What the AI does NOT do:

• Decide which rules to include in the rule set
• Write the appeal-letter templates from scratch
• Modify the dollar amounts on your bill
• Send the appeal letter (you do that)
• Provide legal, medical, or financial advice

Rule maintenance, state-law updates, and the appeal-letter templates are owned by the Audra editorial team. We re-verify the state-law dataset quarterly and after every significant statute change.

We only cite sources you can verify yourself:

• CMS (cms.gov) for NCCI edits, MUE, fee schedules, and Medicare coverage policies.
• HHS (hhs.gov/nosurprises) for No Surprises Act consumer guidance.
• State Attorney General sites — official .gov consumer-complaint portals (links per state at /state/[slug]).
• Your insurer's published policy — pulled from the issuer's public Summary of Benefits + the EOB you upload.

We do NOT cite blog posts, AI-generated summaries, or paywalled secondary sources for rules. If we can't link the source, we don't use the rule.

• State-law dataset — quarterly. Last reviewed: May 21, 2026.
• CMS NCCI edits — quarterly (matches the official CMS release cadence).
• Provider watch-lists — quarterly. See /providers.
• Blog posts — each post carries a visible “Last reviewed” date. Posts that touch federal or state law are re-verified at least once a year, sooner when the underlying statute changes.

Audra is an informational analysis tool, not a law firm, medical provider, or financial advisor. The findings and appeal letters Audra produces are starting points for your own action, not legal advice. For situation-specific guidance, consult a licensed professional.

We are also not affiliated with any hospital system, insurer, or government agency. We take no commission from providers, do not sell user data to insurers, and never share uploaded bills with any third party. Full data handling is documented on the /security page.

Email [email protected] with the specific finding, the source URL you believe supersedes ours, and your reasoning. We respond to every substantive correction and update the dataset within 7 business days when warranted.