2,441,528
Rate rows indexed
ReimburseOS / Methodology
Methodology, sources, and computation.
Every number in a ReimburseOS Snapshot traces to a public dataset, a published formula, and a refresh date. This page documents all three. If you find a gap, write us. We publish corrections.
Methodology version 2026.05 · last updated 2026-05-12
// in plain english
Insurers are required to publish machine-readable contracted rate files. Those files are huge and hard to use. ReimburseOS indexes public payer rate data, matches it to public NPI records, compares contracted rates at the CPT and payer level, and shows where a practice appears below peer benchmarks. No EHR integration. No PHI. No magic.
On this page
// 01Data sources
Four public inputs are composed for every benchmark. Each is independently auditable. Each carries a federal mandate or a federal license. No private aggregator. No scraped panel.
// our dataset
ReimburseOS owns and operates a normalized reimbursement dataset built on public federal filings.
We ingest, parse, and normalize federal Transparency-in-Coverage machine-readable files directly from each payer's CMS-mandated CDN. No private aggregator stands between the raw federal source and our index. TwinFlame Group owns and operates ReimburseOS.
// source 01 — provider identity
CMS NPI Registry (NPPES)
We resolve every input NPI against the federal NPI Registry to confirm the practice name, taxonomy code (specialty), and primary practice address. No PHI. No patient. Just the public provider entry that any U.S. payer is required to use to credential you.
npiregistry.cms.hhs.gov →
// source 02 — contracted rates
CMS Transparency-in-Coverage Machine-Readable Files
Every health insurance issuer in the United States is legally required to publish, monthly, machine-readable files containing the negotiated rate between the issuer and every in-network provider, for every billed item or service. We index the in-network rate files (not the allowed-amount or RxIN files). As of 2026-05-12 we hold 2,441,528 contracted-rate rows across 27 commercial payers and 20 states — the long tail is still being ingested.
CMS Final Rule reference →
// source 03 — medicare benchmark
CMS Physician Fee Schedule (PFS) 2026 + RVU file
For every CPT/HCPCS code we use, we pull the CY2026 PFS Relative Value Units (work, practice expense, malpractice) and the per-locality Geographic Practice Cost Index (GPCI). This produces the locality-adjusted Medicare allowed amount, the floor against which the commercial rate is compared.
CMS PFS lookup →
// source 04 — payer documentation
Public payer fee-schedule + policy bulletins
For modifier handling (e.g., -25, -59, -RT/-LT, -50), bundling rules, and place-of-service multipliers, we use the publicly posted provider manuals of every commercial payer we track. We do not use NCCI edits as price-truth; we use them as bundling-truth.
// 02Computation
For every CPT code returned for your practice, we compute four anchors and one variance.
The four anchors
// per CPT, per payer, per state, per locality
peer_p25 = percentile_cont(0.25) WITHIN GROUP (ORDER BY negotiated_rate)
peer_p50 = percentile_cont(0.50) WITHIN GROUP (ORDER BY negotiated_rate)
peer_p75 = percentile_cont(0.75) WITHIN GROUP (ORDER BY negotiated_rate)
cms_pfs = (work_rvu*work_gpci + pe_rvu*pe_gpci + mp_rvu*mp_gpci) * conversion_factor
The peer band is computed across all NPIs sharing your specialty (taxonomy code) and state. We exclude rates outside the 5th–99th percentile to defend against MRF parsing errors and stale provider-reference IDs.
What "underpayment" means in our model
An underpayment is any CPT/payer pair where your observed contracted rate sits below the peer 50th percentile for your specialty/state. We surface three flavors:
- Mild gap — your rate is 4–10% below peer p50. Negotiable at next contract renewal.
- Material gap — your rate is 10–20% below peer p50. Worth a mid-cycle rate-review letter.
- Severe gap — your rate is >20% below peer p50 or below the CMS PFS floor. Often a coding/contract drift, not negotiation — fixable with a formal amendment request.
// 03Locality (GPCI) adjustment
Commercial rates vary dramatically by metropolitan area — a 99213 in Manhattan is not a 99213 in Modesto. We use the CMS Geographic Practice Cost Index (GPCI) to normalize for cost-of-practice differences. Your practice ZIP is mapped to one of the 112 CMS payment localities, and the resulting GPCI vector (work, practice-expense, malpractice) is applied to the CMS PFS anchor.
// example: CA-Sacramento (locality 99, GPCI 1.041) vs CA-Other (locality 75, GPCI 1.000)
98941 cms_pfs_sacramento = $36.18
98941 cms_pfs_ca_other = $34.76 // 4.1% locality delta
// 04Refresh cadence
Rates are updated monthly. Every health insurance issuer must publish a refreshed in-network MRF on the first business day of each month (CMS-9915-F §149.220(d)). Our ingestion crawler picks up the new files within 48 hours of publication, re-runs the parse, and updates the rate index. The last successful payer ingest before this page was rendered: 2026-05-11.
- Daily — NPI Registry deltas (new providers, address changes, deactivations)
- Monthly — TiC MRF rate refresh per payer
- Annually — CMS PFS conversion factor + GPCI vector + RVU file (CY-rollover, every November)
// 05Confidence logic
Every benchmark we surface carries a confidence label so you know how much weight to put on it. The label is driven by row count behind the peer band, not opinion.
- High — 150+ rate rows behind the peer band for that CPT, payer, state, and specialty cohort.
- Moderate — 25 to 149 rate rows behind the peer band. Directionally useful, narrower sample.
- Low — 1 to 24 rate rows. Surface only with explicit "thin sample" labeling.
- Processing — payer ingestion in flight. Shown only with a "still ingesting" badge.
- Unavailable — no public TiC rows yet for that CPT and payer combination in your state. We show this instead of guessing.
// 06Limits and what we don't claim
The benchmark layer is precise about contracted rates. It does not, and cannot, know everything about how a practice runs. We surface what we have. We label what we don't.
- We report documented reimbursement opportunity, not realized recovery. Negotiation outcomes depend on you, your payer relationships, and your contract terms.
- We surface contracted rates from public MRF filings. We do not have visibility into your remittance advice unless you choose a paid workflow that supports it.
- We do not know your patient mix, your internal billing realities, your modifier patterns at the encounter level, or payer-specific authorization rules.
- Some payers publish incomplete MRFs (missing modifier rows, missing place-of-service variants, sparse provider-reference resolution). When we can't confidently match your NPI, we say so in the Snapshot rather than guessing.
- We do not have every payer in every state yet. 27 payers across 20 states as of 2026-05-12. The long tail (regional plans, Medicare Advantage, state Medicaid MCOs) is being ingested.
How to interpret a result
A gap is a starting point, not a verdict. If a CPT and payer pair shows up below peer p50, the right next move is to compare that rate against your actual contract language, your fee schedule attachment, and your last renewal letter. The Snapshot tells you where to look. The Practice Audit and counteroffer memo turn that into a negotiation package.
What ReimburseOS does NOT know
- Anything about your patients.
- Your real-world payer mix or volume by code.
- Your contract clauses (carve-outs, multi-procedure reductions, modifier policies) unless you submit them on a paid tier.
- Whether a payer will agree to renegotiate. We surface leverage. We don't promise outcome.
Snapshot delivery
Results appear in your browser in about 15 seconds. No email required. No credit card. No PHI.
// 07Corrections policy
If you can show that a rate in our Snapshot is wrong — wrong CPT, wrong payer, wrong locality, stale MRF version — write david@reimburseos.com. We will re-run the parse, post the correction, and credit any paid Snapshot affected by the error. No NDA. No PR cycle. Methodology is the moat.
The methodology is open. The data is public. The compute is yours.
Type your NPI. Inspect the anchors. Audit the variance. Then decide what to negotiate.
⚡ Run My Free SnapshotMethodology v2026.05 · refreshed monthly