Methodology

In One Paragraph

For each line item we recognize on your bill, we cross-check the charge against four authoritative public-data sources: Medicare reimbursement rates (CMS Physician Fee Schedule), the National Average Drug Acquisition Cost for pharmacy charges, your specific hospital’s published price-transparency rates (cash-pay and insurance-negotiated, when available), and CMS National Correct Coding Initiative edits for bundling violations. We also resolve the facility through the NPPES NPI Registry to surface hospital-specific rates and identify nonprofit (§501(c)(3)) status. We surface gaps between billed amounts and these benchmarks so you can evaluate whether a charge is in line with what providers commonly accept. These are estimates, not authoritative prices, and they are not a guarantee that any charge is wrong.

Four Layered Data Sources

We layer four federal / authoritative datasets, each grounding a different part of the audit:

1. CMS Physician Fee Schedule (PFS) — Medicare reimbursement rates for roughly 10,000 CPT and HCPCS procedure codes, published by the Centers for Medicare & Medicaid Services. Annual updates each January with periodic interim revisions. This is the legal floor for what Medicare pays. Source: cms.gov physician-fee-schedule.
2. NADAC (National Average Drug Acquisition Cost) — pharmacy acquisition cost for 33,000+ drug NDCs, sourced weeklyfrom a CMS survey of roughly 60,000 retail pharmacies. We use NADAC to evaluate IV fluids, J-code injectables, and pharmacy line items, where chargemaster markups commonly run 50–500× over the manufacturer-acquisition cost. Source: medicaid.gov NADAC.
3. Hospital Price Transparency (HPT) files — federally-mandated cash-pay and insurance-negotiated rates per facility, published under 45 CFR §180.50. Every U.S. hospital must post a machine-readable file showing what it actually accepts from self-pay patients and the median rate it has negotiated with major insurers for each service. When your facility’s HPT data is in our ingested dataset, we use those hospital-specific rates as the cash-pay and commercial benchmarks. When it is not, we fall back to Medicare-anchored estimates and label them as such in the line-item table. Source: cms.gov hospital-price-transparency.
4. CMS National Correct Coding Initiative (NCCI) Procedure-to-Procedure edits — the official rules for which CPT codes are allowed to be billed together. Quarterly updates. We auto-flag unbundling violations (for example, billing a comprehensive metabolic panel and the individual component panels on the same date of service). Source: cms.gov ncci-edits.

Facility Identity (NPPES NPI Registry)

When the bill names a facility, we resolve its name and address against the NPPES NPI Registry (the public registry of all U.S. healthcare providers maintained by CMS). Resolving the National Provider Identifier (NPI) lets us look up the hospital’s specific HPT file (rather than relying on a national average), and lets us identify whether the facility is an IRS §501(c)(3) nonprofit — relevant because nonprofit hospitals are required by ACA §501(r) to have a published Financial Assistance Policy.

Three-Benchmark Display

For each line item with a CPT code, we show three benchmarks side-by-side in the audit results table:

• Cash-pay— what the hospital accepts from self-pay patients (from its HPT file when public data is available; Medicare-anchored estimate at roughly 1.0–1.2× Medicare otherwise).
• Insurance pays— median negotiated rate insurers actually pay this hospital for this service (from its HPT file when available; Medicare × 1.8 estimate otherwise).
• Medicare— CMS Physician Fee Schedule national-average benchmark, GPCI-adjusted by state when the bill identifies the facility’s location.

The bigger the gap between what you were billed and these benchmarks, the stronger your dispute.

Locality Adjustment (GPCI)

Medicare rates vary by geographic region under the Geographic Practice Cost Indices (GPCI). High-cost metros run roughly 1.20–1.40× the national base; rural areas run roughly 0.85–0.95×. When we can identify the facility’s state from the printed address, we label our estimates accordingly (“approximate for [STATE] locality, GPCI-adjusted”). When the state is not visible on the bill, we show the national-average rate and label it as such. We do not fabricate precise locality-adjusted numbers.

Commercial Multiplier (Fallback)

When a hospital’s specific HPT data is not yet in our system, we fall back to a Medicare-anchored commercial estimate. Commercial fair-market multipliers typically run 1.5–2.5× the Medicare rate, with high-cost states (CA, NY, MA) often at the top of that range and rural Midwest / Deep South often at the bottom. We use a 1.8× midpoint when no hospital-specific data is available, and label it as an estimate. These are not contractual rates and are not a prediction of what your specific provider has agreed to with your specific insurer.

What the AI Looks For

Powered by Google Gemini 2.5 Flash (paid tier), the analysis screens uploaded bills for patterns commonly associated with potential billing errors:

• Emergency-department visits coded at higher acuity levels than the documented care suggests (potential upcoding)
• Charges that may have been unbundled, cross-checked against the NCCI Procedure-to-Procedure edits dataset
• Line items that may be duplicates (same service charged twice, sometimes with differently-worded descriptions)
• Charges for items that may not have been provided (phantom charges)
• Supplies and pharmaceuticals priced significantly above public-data benchmarks (HPT cash-pay, NADAC drug acquisition cost)
• Summary-only bills that lack itemization needed for verification

The system surfaces these for your review. It does not adjudicate, does not tell you whether any charge is right or wrong, and does not communicate with your provider on your behalf.

AI Provider

Bills are sent to Google Gemini (paid tier) for analysis. The paid tier is contractually prohibited from using your inputs to train models. Bills are processed in memory and not stored on our servers. See our Privacy Policy and Consumer Health Data Privacy Policy for details.

Limitations

• Estimates only. None of the four data sources is a market-clearing price for your specific bill — CMS PFS is a Medicare reimbursement reference, NADAC is a pharmacy-acquisition cost, HPT is the hospital’s self-published rate (which providers do sometimes dispute), and NCCI edits define which codes can be billed together but not what a service should cost.
• AI output can miss real issues and can flag patterns that are not actually problems. You must review every flagged item before taking action.
• Hospital-specific HPT data is not yet ingested for every U.S. hospital. Where it is missing, we fall back to Medicare-anchored estimates and label them as such in the results.
• We do not yet publish customer outcome data. Whether a hospital adjusts a charge depends on your bill, your provider, and the action you choose to take. We do not predict your result.
• The Service is informational self-help software. It is not legal, medical, or billing advice, and it is not a substitute for a credentialed medical billing advocate, attorney, or CPA.
• We cannot guarantee that any specific charge is wrong, that any specific letter template will produce a reduction, or that any specific outcome will result.

Update Cadence

• CMS Physician Fee Schedule: annual update each January, with periodic interim revisions. We refresh our reference table at each release.
• NADAC: CMS publishes weekly drug-cost updates; we ingest periodically to keep pharmacy benchmarks current.
• Hospital Price Transparency: hospitals control their own publication cadence; most update at least annually. We re-ingest from public sources on a rolling schedule.
• NCCI edits: CMS publishes quarterly updates; we refresh on the same cadence.
• NPPES NPI Registry: queried in real time per analysis, with results cached for 30 days to limit API traffic.

The methodology used for each analysis reflects the data available on the date you ran the analysis.

Questions

Questions about our methodology, sources, or how we estimate a specific figure? support@medibillsaver.com