Methodology

Under the federal Transparency in Coverage rule (45 CFR 147.212), health plans must publish machine-readable files listing the rates they have negotiated with in-network providers. Where a plan includes dental coverage, those files carry dental procedure codes and a negotiated dollar amount for each contracted provider group.

We stream those files directly from the payers, keep only the 41 treatment codes people actually search for, and discard the rest. Every file we read is listed with its publication date.

The unit of observation is one provider group — in practice, one dental practice — not one dentist and not one rate row.

This matters more than it sounds. A practice usually appears in these files several times for the same treatment, holding different rates for different plan tiers. Counting each as a separate data point would inflate the apparent sample size roughly twentyfold and make thin data look robust. So we collapse every rate a practice holds for a treatment to that practice's median, and the practice contributes exactly one observation.

Payer files identify providers by NPI, with no address. We resolve each provider group to a practice location through two public CMS sources, in this order:

• NPPES, the national provider registry, which gives a city and a state. 38,132 groups resolved this way.
• Medicare Fee-For-Service Public Provider Enrollment, which gives a state but no city. 3,831 groups resolved this way.

The distinction matters and is enforced in the data: a group located only to a state contributes to state and national figures but can never appear in a city figure. Every city page on this site is built purely from city-resolved practices.

Groups we cannot resolve to an address at all are excluded, not guessed. 41,963 groups were located in total.

These files contain placeholder and administrative values. In testing we found a routine check-up carrying rates of both $3 and $10,000. Neither is a real price.

So we drop observations outside the 2nd to 98th percentile of the national distribution for that specific treatment, and publish the count: 37,640 of 1,040,653 observations were removed this way. The 1,003,013 that remain are what the site is built on.

A city figure is published only when at least 8 distinct practices support it. A state figure needs at least 15. Below those thresholds we show nothing.

This is a deliberate trade. Some cities have no page and some treatments are missing from a state that has others. We would rather have gaps than publish a national average dressed up as a local price.

This is the most important limitation on the site, which is why it appears on every page. A negotiated rate is what a dental plan pays a practice it holds a contract with. If you have no insurance and walk in off the street, that is not the price you will be offered — uninsured patients are generally quoted at or above the highest negotiated tier.

We do not have observed cash prices and will not model one and present it as data. Where you see an estimated self-pay range, it is the upper half of the same observed negotiated distribution, labelled as an estimate everywhere it appears. It is a budgeting aid. The only reliable cash price is a written quote.

• Standalone dental plans are exempt. The rule treats standalone dental cover as an excepted benefit, so dedicated dental carriers publish nothing. Our data is dental embedded in medical plans. That is a real selection effect, not the whole market.
• One payer network, so far. Every figure on this site currently comes from a single payer's embedded-dental network. That payer republishes the same national dataset under dozens of legal entity names; we verified the files are byte-identical and extract it once rather than counting it repeatedly. It does mean these are one insurer's negotiated rates, not a market-wide consensus, and a practice outside that network does not appear at all. Adding further payers is the main improvement still outstanding.
• Coverage is uneven. Metros with many contracted practices produce solid samples; rural areas often fall below the publication threshold.
• Negotiated rates move. Payers republish monthly; every figure carries the date it was built.
• No cash-price survey. The commercial dental fee surveys covering self-pay pricing are licensed products we have chosen not to use, so we have no observed cash data at all.

If a figure here looks wrong, it may be. The dataset is downloadable and every source file is listed, so anyone can check the arithmetic.