How to check whether a US hospital or clinic is still operating

Before you ship to a US hospital, sign it as a distributor, or put it in a diligence memo, confirm one thing first: that it is still on the official register. Being named in the file is not the same as operating — more than half the rows in this file are historical.

Published 2026-10-22Data CMS 2026 Q2Read 7 min

A complete-looking file where 55.7% of rows are history

The Q2 2026 clinical laboratory file has 681,059 rows. Filter on the termination code and 301,760 remain in force — 44.3%. The other 379,299 are terminated, with addresses, certificate numbers and certificate types all fully populated.

The number itself is not the problem. What it looks like in practice is:

So judge a record by its status field, not by how full it is. For someone building a list that is a filter; for someone doing diligence it is a lead — a terminated row says the entity re-certified, changed hands, or genuinely exited.

Two files, two decay rates

Most people treat "US hospitals" and "US labs" as one dataset and apply one cleaning rule. That is the commonest mistake here, because the two files have completely different life structures:

The lab file decays 3.4 times faster. In practice: a 20,000-row list of US labs, left uncleaned for a year, loses about 1,440 live rows. A hospital list of the same size loses roughly 420.

Why the gap? Because the two register different things. A lab holds a certificate, which expires and is renewed — and each renewal leaves a termination behind. A facility registers for Medicare enrolment once, and it stays valid until it actually exits. So history piles up faster in the lab file.

The 7.2% is an approximation: 5.5 years of terminations spread over the current live count. It measures how many rows in a list expire per year, not how many of the same organisations churned — renewals also produce terminations.

Three free official sources

The first two are the same source family, published quarterly on data.cms.gov — and they are the source of every figure here.

A one-minute check, in three steps

1. Resolve the entity before reading status. Duplicate names are common. Match on address plus certificate number (CCN or CLIA), never on company name alone.
2. Read the termination code, then the date. 00 is in force; anything else is terminated, and the date tells you whether that was three months or twenty years ago.
3. Cross-check the state licence. Still in force? Confirm the state licence. Terminated? Check whether it re-registered under a new number.

Step 3 is the one people skip, because "terminated" reads like a verdict. The data does not support that reading: 31.9% of terminated records held their certificate for less than two years, and the median tenure is four years. Certificates are short-lived instruments. Treating every termination as a closure badly overstates exits.

These three steps are really one table: entity, certificate number, status test, termination date, source line. Fill those five and the verdict follows.

Same rule, twenty points apart by state

Break the live share out by state and the spread is wider than expected: New York 36.7%, California 38.7% and Illinois 39.1% at the bottom; Nebraska 56.7%, Minnesota 56.4% and Idaho 55.6% at the top. Almost 20 points between the extremes.

Nothing in this file explains why — it records certificate status, not state licensing policy, clinic density or test mix. So we leave one practical warning: using a national average decay rate to estimate the quality of one state's list will read either too optimistically or too pessimistically. If you build lists by state, recompute by state.

FAQ

Also in this series: the ten drugs Medicare spends most on · who supplies SpaceX