Last month 2,733 inspections left CMS’s monthly file.
They are still shown here. This is what dropped out of the monthly file between CMS's 26 August 2026 release and its 8 October 2026 one.
Data as of , released by CMS on October 8, 2026. Source: CMS Provider Data Catalog.
Surveys present in the previous file and absent from the next.
753 of them were written at actual harm or above.
16.6% of the 14,690 nursing homes CMS currently lists.
How it happens
Two mechanisms, and the second one is easy to miss
CMS republishes its Health Deficiencies file monthly, and it is destructive: it carries roughly the last three inspection cycles per facility, so each release discards what has fallen out of that window. Nothing about this is hidden — it is a stated retention policy — but the effect is that the monthly file, and Care Compare which is built on it, shows less than a nursing home’s full history. The older findings are not gone: CMS keeps them in its quarterly bulk downloads, which is where the history on this site comes from.
2,733 inspections were dropped outright this month, taking 10,893 findings with them. Separately, 98 inspections CMS still publishes lost 338 findings between the two releases — the survey is still there, and part of what the surveyor wrote is not. That second mechanism leaves no trace at all in the current file.
In the same month 2,412 new inspections and 10,610 new findings arrived. The file is not shrinking; it is a moving window.
What went
A cohort leaves at once, and you can watch the edge move
Removal is not a gradual decay by age. Inspections from 2023 and earlier lost 10.6% of what we hold this month. Everything from 2024 onward lost 0.0% — 28 inspections out of 70,377. The window has an edge, and this release advanced it.
It is sharper than a year makes it look. 1,729 of the 2,733 removals are dated August 2023 alone — 68.2% of that month, against 0.9% of the month before and 0.7% of the month after. A cohort of inspections drops together because retention works on inspection cycles, not on dates.
Which is also why a few recent ones go. The newest inspection removed this month was conducted on 2 July 2026. 16 of the removals were less than a year old, and 15 of those were complaint investigations — a complaint is filed against the cycle it falls in, so its date interleaves with newer ones.
- 20175of 35 held
- 201851of 422 held
- 2019153of 1,841 held
- 202050of 694 held
- 2021126of 1,894 held
- 2022264of 3,822 held
- 20232,056of 16,865 held
- 202412of 27,881 held
- 20251of 26,150 held
- 202615of 16,346 held
Bars are counts of inspections removed, scaled to the largest year. The “held” figure is what this site holds for that year from CMS’s monthly files, not everything that ever existed — earlier years had already been thinned by earlier releases.
By harm level
Removal is even across the severity ladder
Between 2.5% and 4.1% of citations were removed at every level of the scale. There is no sign that severe findings are dropped more readily than minor ones — the window is blind to severity, which is precisely why it takes serious findings with it. 489 of the removed inspections contained at least one finding at actual harm or above.
| Level | Removed | Still published | Share removed |
|---|---|---|---|
| L | 21 | 607 | 3.34% |
| K | 86 | 2,004 | 4.11% |
| J | 227 | 6,637 | 3.31% |
| I | 2 | 19 | 9.52% |
| H | 19 | 439 | 4.15% |
| G | 398 | 12,858 | 3.00% |
| F | 746 | 28,988 | 2.51% |
| E | 2,467 | 93,485 | 2.57% |
| D | 6,896 | 264,694 | 2.54% |
| C | 119 | 4,402 | 2.63% |
| B | 181 | 4,814 | 3.62% |
The range quoted above covers levels with at least 100 citations on record. Level I holds too few nationally for a share to mean anything, and is shown in the table regardless.
What it is not
How to read this page
This is one month, not a total. It is the difference between two consecutive releases of one file. This site as a whole holds a great deal more than the current monthly file, but that figure is arrived at differently and the two should not be added together.
Removal is not a judgement about a facility. An inspection leaving the file says nothing about what was found in it. 5 facilities lost their entire published health-survey history this month, which means anyone reading Care Compare now sees no findings for them at all.
Change is measured between snapshots, never between survey years. Survey years contain different populations of facilities, so a year-on-year comparison of counts measures who was surveyed rather than what changed.
Deficiency Index is independent and is not affiliated with, endorsed by, or approved by CMS, Medicare, or any government agency. About this data · Report a correction