CMS found errors in half of directory listings. Here is the part we can prove.
Starting in 2016, CMS sent reviewers to call doctors listed in Medicare Advantage directories. Nearly half the listings failed.
That finding gets quoted constantly. It deserves a closer look, and so do our own numbers.
The claim
CMS reviewed 54 plans in its first round. Reviewers found 45.1% of listed locations wrong in some way. The second round, covering 64 plans, came in at 52.2%. A third round landed near 49%.
Wrong meant something concrete. A phone number that did not work. A doctor no longer at the address. A listing marked "accepting new patients" when the office was not.
The last of those reviews is years old now. The stakes have grown since.
KFF counts 34.1 million people in Medicare Advantage in 2025. That is 54% of eligible beneficiaries, up from 19% in 2007. Each of them may pick a doctor from one.
The check
We compare what payer directories publish against the provider's own record. Our scans read each directory on a recurring schedule, and every read carries a date.
Here is what the disagreements look like, by type, as of 14 November 2026.
Big numbers. Too big to quote without a caveat, honestly.
One row in our table is not one problem. The same disagreement can appear across several directory files. So we count twice, and show both.
| What we count | Number |
| --- | --- |
| Rows in our mismatch table | 2,366,143 |
| Distinct disagreements behind them | 1,325,942 |
Source: Argoseer scans, Nov 2026 (draft: sample is mostly Texas)
The distinct count is the one that matters. It is still more than a million.
What one disagreement looks like
Take a family physician who joins a new group across town. She updates her NPPES record the same week. Her old clinic keeps her name on its roster for months.
One plan's directory now shows two addresses for her. Another shows only the old one. A third lists the new address under the wrong specialty.
A patient picks the old address because it is closer. The front desk there has never heard of her.
Nothing in that chain is fraud. Each system did what it was built to do. The records simply stopped agreeing, and nobody owned the gap.
That is the pattern behind the "wrong location" bar above. It is also why our first comparison is with NPPES and the practice's roster. A directory can only be wrong relative to something.
The verdict, with a confession
Now the uncomfortable part. Take one finding: a provider missing from a payer directory.
A single detector flags it far more often than two detectors agree on it. We publish only the confirmed figure.
About 18% of those findings clear the higher bar. That is 304,440 confirmed, against 1,345,998 that rest on one signal.
Single signals are not useless. Many turn out right once a second source catches up. Some are artefacts: a directory file that dropped a page, a name spelled two ways.
We cannot tell those apart from one read. So the single-signal pile stays a lead, never a claim.
Why say the smaller number out loud? Because a practice manager acting on our data deserves the defensible version.
A missed listing costs a patient a wasted call. A false alarm costs a credentialing team an afternoon. Both are real costs.
So where does that leave the CMS finding? Mostly intact. Directories still disagree with provider records at scale. What changed is the method: we can show the disagreement with a date and two sources, not a phone survey.
Our sample leans heavily on Texas today. We label it draft until that validation is done. Other states are being added.
If you run a practice, the useful question is narrow. Which of your own providers are visible, and in which plans? We built the practice view for that.
The harder question belongs to the plans. If two checks agree a doctor is missing, how long should that listing stay wrong?
Catch the mismatch before a payer does
Argoseer watches your providers' records across NPPES and payer directories and tells you what changed, so a stale address or taxonomy gets fixed before it turns into a denial.
Join the pilot →Argoseer
Building the future of provider data intelligence.
