How the Index is calculated
One number for how much wrong a directory is carrying: severity-weighted defects per 1,000 listings, over the classes actually checked, at a stated method version. It is a defect-burden score, not an accuracy score. We never publish a number that implies the rest of the directory is correct.
HOW IT IS COMPUTED
location = weight of its single worst defect (CMS does not stack them) index = Σ location ÷ listings × 1,000
The worst-defect rule only bites when one location carries two defects. We do not yet hold a location identifier linking the two classes we count, so each is treated as its own location, which is the assumption that makes the numbers below upper bounds. Any overlap moves an index down, never up. Said plainly because it is an assumption and not a measurement.
CMS's weights, and what we can see
CMS ran these reviews by telephone across 10,504 provider locations. We run them against the federal registries, which sees some of the same defects and cannot see others. Both columns are shown, because a weight we cannot yet detect is not a weight we get to claim.
| CMS deficiency | Weight | How we detect it | Source | Status |
|---|---|---|---|---|
| Provider should not be listed at any listed location | 3 | Deactivated, retired or deceased NPI | NPPES | SCORED |
| Provider should not be listed at this location | 3 | Moved, still listed at the old address | NPPES address history | SCORED |
| Phone number needs updating | 3 | Listed phone contradicts the registry | NPPES / PECOS | NOT YET RUN |
| Provider is NOT accepting new patients | 3 | Not in any registry. Needs outreach | Resolve agents | NOT REGISTRY-DERIVABLE |
| Address needs updating | 2 | Listed address contradicts the registry | NPPES | NOT YET RUN |
| Address (suite number) needs updating | 1 | Street matches, suite does not | NPPES | NOT YET RUN |
| Provider IS accepting new patients | 1 | Not in any registry. Needs outreach | Resolve agents | NOT REGISTRY-DERIVABLE |
| Specialty needs updating | 1 | Listed taxonomy contradicts the registry | NPPES taxonomy | NOT YET RUN |
| Provider name needs updating | 0 | CMS scores this zero: not a barrier to care | NPPES | NOT YET RUN |
Three rules that keep the number honest
- The weights are CMS's, not ours. We do not tune them. Any argument for a different weight is an argument with the regulator, and it is not one we would win in front of a buyer. When CMS revises them, we revise and mark a break in the series.
- Coverage is part of the score. Every score is published as "N of 9 CMS deficiency types checked". Two plans compare only on the same set. An unchecked type is never scored as a zero.
- Our number is not CMS's number. CMS telephones a sample of listings and reports a percentage of its own maximum. We diff registries over a different sample and report points per 1,000 listings. The scale is deliberately different so the two are never mistaken for each other.
Scored so far
| Plan | Listings reconciled | Index | Coverage | Status |
|---|---|---|---|---|
| Plan B national MA | 6,000 | 20.0 | 2 of 9 | VERIFIED 2026-08-06 |
| Plan A national MA | 5,000 | 11.4 | 2 of 9 | VERIFIED 2026-08-06 |
| Your plan | Not yet reconciledRequest your proof file |
Method v0.2: automated diff of each plan's public FHIR Plan-Net provider directory against NPPES, PECOS and OIG LEIE, weighted by CMS Table 3 (Online Provider Directory Review, Round 3, 28 November 2018). Every score is a floor over the deficiency types actually checked, not a full accuracy audit. v0.1 scored on our own weights; those numbers are withdrawn, not converted. Plans are anonymized; no plan is named without its permission.