Retired or just unconfirmed? Reading a doctor list
Not every mismatch in a doctor list is an error. We checked 1,909,677 insurer listings against the federal provider registry. We found 5,478 providers who had already been retired before we captured the list. A mismatched address is different. It is usually unconfirmed, not wrong. Treat the two very differently.
The short version
Your insurer's doctor list can disagree with the federal registry in two ways.
Some disagreements are hard facts. The registry says the provider's ID number was retired. A listing like that is wrong.
Other disagreements are soft. The address in the list is not the one on file. That usually means "not confirmed", not "wrong".
This post shows how big each group is in our data.
What we compared
Marketplace insurers publish their doctor lists as machine-readable files. We captured 1,909,677 provider listings from those files between 26 and 28 July 2026.
We then compared each listing against NPPES. NPPES is the federal registry that issues every provider's National Provider Identifier, or NPI.
We ran the comparison three times. The most recent run was on 15 September 2026.
Two kinds of mismatch
Here is the split that matters most.
| Kind | What the registry says | What it means for you |
|---|---|---|
| Hard defect | The NPI was retired, does not exist, or belongs to a different type of provider | The listing is wrong |
| Soft signal | The name or address differs from the registry | The listing is unconfirmed |
A hard defect is not a matter of opinion. A soft signal needs a phone call before you trust it or dismiss it.
The hard defects
These are counts from our 15 September 2026 run, across all 1,909,677 listings.
| Hard defect | Listings | Share of all listings |
|---|---|---|
| Provider's NPI has been retired | 5,996 | 0.31% |
| Listing names an individual, registry says organization (or reverse) | 7,167 | 0.38% |
| NPI does not exist in the registry at all | 62 | under 0.01% |
That is 13,225 listings that the registry flatly contradicts.
As a share, it is small. For the person who books with one of them, it is the whole story.
The soft signals
Soft signals are much bigger. They are also much weaker evidence.
| Soft signal | Listings | Out of | Share |
|---|---|---|---|
| Address not found in the registry | 1,201,623 | 1,903,633 checked | 63.12% |
| Address matches the registry | 702,010 | 1,903,633 checked | 36.88% |
| Name does not match the registry | 32,435 | 1,552,447 checked | 2.09% |
That 63.12% figure looks alarming. It should not be read as an error rate.
Why an unmatched address is not an error
The federal registry stores one main practice location per provider.
Many doctors work in more than one place. A surgeon may see patients at a clinic and operate at a hospital. Only one of those can match the registry.
So an address the registry does not have is simply unconfirmed. We will not call it wrong, because the evidence does not support that.
Name mismatches are similar. In one sample, the list used a short first name. The registry had a much longer full name. That may well be the same person.
Retired before the list was even captured
This is the strongest finding in the data.
Of the 5,996 retired NPIs in our latest run, 5,478 had been retired before we captured the directory. The insurers were still listing them in late July.
| Retired before we captured the list | Listings |
|---|---|
| Total | 5,478 |
| Retired more than a year earlier | 3,010 |
| Retired more than five years earlier | 312 |
| Retired within the prior 90 days | 680 |
The median gap was 466 days. Half of these listings named a provider retired well over a year earlier.
The oldest retirement we found dates to May 2009.
Who the retired listings are
Most of them are not individual doctors.
| Provider type | Retired before capture | Share |
|---|---|---|
| Facility | 3,696 | 67.47% |
| Individual clinician | 1,502 | 27.42% |
| Group practice | 280 | 5.11% |
Facilities make up two thirds. We do not know why from this data, and we will not guess.
It does mean a hospital, lab or clinic listing deserves the same check as a doctor listing.
What happens to a list nobody updates
Last month we said two data points were not a trend. We now have three.
The directory rows stayed exactly the same across all three runs. Only the registry moved. So this measures how a frozen list decays.
| Run date | Retired NPIs still listed | Name mismatches |
|---|---|---|
| 1 August 2026 | 5,337 | 30,753 |
| 15 August 2026 | 5,630 | 31,426 |
| 15 September 2026 | 5,996 | 32,435 |
Retired listings rose by 659 in 45 days, or 12.35%. They rose in both intervals.
This does not describe the insurers' live directories. It describes our July copy of them. It shows what a list goes through when no one touches it.
What the rules require
Insurers are supposed to keep these lists current.
For marketplace plans, 45 CFR 156.230 requires an "up-to-date, accurate, and complete provider directory". It must be viewable without an account.
The No Surprises Act provision at 42 U.S.C. 300gg-115 goes further. Plans must verify directory information "not less frequently than once every 90 days". They must also have a process to remove providers they cannot verify.
The registry side updates too. CMS publishes a monthly NPI deactivation file. That is the file we check against.
We are not lawyers. We are reporting what the data shows, not whether any insurer broke a rule.
How to check a listing yourself
You can run a smaller version of our check in a few minutes.
Look up the provider's NPI in the federal NPI Registry. If no active record comes back, treat the listing with real caution.
If the NPI is active but the address differs, call the office. Ask whether they see patients at the listed address. Ask whether they take your specific plan.
Write down the date and who you spoke to. That record helps if a bill comes back out of network.
Hard defect or soft signal: what to do
| What you find | How strong the evidence is | What to do |
|---|---|---|
| NPI retired | Strong | Treat the listing as wrong. Find another provider. |
| NPI missing from registry | Strong | Treat the listing as wrong. |
| Wrong provider type | Strong | Check you have the right provider. |
| Address differs | Weak | Call and confirm the location. |
| Name differs | Weak | Check it is the same person. |
What this does not tell you
We want to be plain about the limits.
We check listings against one federal registry. An active NPI does not prove a provider takes your plan. It only proves the ID is live.
Our directory copy is from July 2026. Insurers may have fixed some listings since then.
And a retired NPI is strong evidence a listing is stale. It is not always proof the provider stopped seeing patients. We cannot see why any NPI was retired.
Where these numbers come from
Every figure in this post comes from our own data, queried live on the day of publication.
- Directory source: 1,909,677 provider listings from insurers' machine-readable directory files, captured 26 to 28 July 2026
- Registry source: the federal NPPES registry and its monthly deactivation file
- Comparison runs: 1 August, 15 August and 15 September 2026
- Retirement timing: each retired listing's deactivation date, compared against the 26 July 2026 capture date
Addresses the registry cannot confirm are reported as unconfirmed, not as errors. External requirements are cited to 45 CFR 156.230, 42 U.S.C. 300gg-115 and the CMS NPPES data files page.
Questions people ask
How many insurer listings name a retired provider?
In our 15 September 2026 run, 5,996 of 1,909,677 listings named a provider whose NPI had been retired. Of those, 5,478 were already retired before we captured the directory in late July.
Is a mismatched address in a doctor list an error?
Usually not. The federal registry stores one main practice location. Many providers work at several. An address the registry does not have is unconfirmed, so call the office to check it.
How long do retired providers stay listed?
Among listings retired before our capture, the median gap was 466 days. That means half had been retired for well over a year. Of the 5,478, a total of 3,010 were retired more than a year earlier.
Are retired listings mostly doctors?
No. Facilities make up 3,696 of the 5,478 listings retired before capture. That is 67.47%. Individual clinicians are 1,502 and group practices are 280.
Does an active NPI mean the provider takes my plan?
No. An active NPI only shows the provider's ID is live. It does not show which plans they accept. Call the office and name your exact plan.
How often must insurers verify their doctor lists?
Under 42 U.S.C. 300gg-115, plans must verify directory information at least once every 90 days. They must also have a process for removing providers they cannot verify.
What share of listings have a hard defect?
It is 0.69% of all listings. Retired NPIs are 0.31%, wrong provider types are 0.38%, and unknown NPIs are under 0.01%. Together that is 13,225 listings.
Is directory accuracy getting worse?
We cannot say that about live directories. Our three runs used the same July copy. Retired listings in that copy grew from 5,337 to 5,996 in 45 days. That shows how a frozen list decays.