An investigation into CNAS data · 2025
How full
is a hospital?
For 198 hospitals, Romania’s published occupancy figure can be reproduced using a 290-day denominator. A full year changes the result. The recorded care stays the same.
The same 11,776 patient-days. The same 40 contracted beds.
The calendar calculation assumes 40 beds available throughout the year. The data do not establish that continuity.
Follow the calculationIn Nucet, 101.52% does not, by itself, prove overcrowding.
On Romania’s public hospital map, a percentage appears to describe a building directly. How full it is. How much room remains. When the national health insurance house, CNAS, launched the map on 17 July 2026, it translated an acute-care average of about 63% into the image of one unoccupied bed in three. The official announcement invites a literal reading.
But another measure appears inside the data. For 198 of the 337 hospitals with all three positive values needed for a check, the percentage can be reconstructed by dividing patient-days by contracted beds and by 290 days. The match is tested within 0.005 percentage points, half the final decimal place of a percentage displayed to two decimal places.
The number 290 has an administrative basis. The rules use it when establishing capacity for contracting. When the resulting percentage is simply labelled “occupancy”, a reader can give the number a meaning that its formula does not support. This investigation follows that difference, hospital by hospital.
NUCET · 2025
Lengthen the year. Keep the recorded care fixed.
For Nucet, the file reports 11,776 acute-care patient-days and 40 contracted beds. Dividing gives 294.4 days of care for each bed in that count. The next step depends on the duration used for comparison.
294.4 days of care per bed
One unit represents one equivalent capacity-day per bed. The arrangement compresses an annual total; it does not show admission dates.
365 days is a constant-capacity assumption, not a certified correction of the CNAS data.
290 days is a capacity convention.
The implementing rules for the framework contract, consolidated through August 2025, specify a bed-use index for establishing hospital capacity. They set 290 days for acute care, 320 for chronic and palliative care, and 360 for the compulsory-admission and long-term-care categories described in the rule. Annex 23, Article 4, PDF page 471 states the values and their purpose.
Reaching 100% of the 290-day reference therefore corresponds to 79.45% in a calculation with the same capacity held constant over 365 days. The 75 days between these references are a difference in denominators. They are not 75 days when we know the hospital was closed.
In Nucet, 294.4 equivalent days exceed the 290-day reference. That produces a result above 100%. Divided by 365, the same data give 80.66%. Neither operation, on its own, describes the pressure on a Tuesday morning or occupancy in a particular ward.
198 matches. 134 unanswered questions.
We checked all 372 records in the file. For acute care, 337 have positive patient-days, contracted beds and a published percentage. Each mark below is one hospital, in the same order as the downloadable register.
- 198 Reproduces 290 days
- 5 Reproduces 365 days
- 134 Reproduces neither
- 35 Insufficient positive data
Five records reproduce the 365-day formula at the same tolerance. Another 134 reproduce neither the 290-day nor the 365-day formula. We have not assigned an explanation to them. Changing bed counts, different reporting bases or an error could contribute; the file alone cannot distinguish them.
For chronic care, a separate check finds 152 of 292 testable records reproducing the 320-day reference. This is another trace of the capacity convention. That check is not included in the 198 acute-care matches.
Which year reproduces your hospital’s percentage?
Search for a hospital or county. Inspect the published inputs and both calculations with a constant bed count. A numerical match supports a hypothesis about the formula; it does not establish the history of available capacity.
Loading the register…
What an average cannot tell you about a bed.
Actual calendar-year occupancy needs the sum of available capacity on each day. If a ward closed for construction, beds moved or the contracted count changed, multiplying one count by 365 does not reconstruct the year. The public file does not provide the daily history needed to do that.
There is another limit even with a perfect record. An annual average can conceal peaks, differences between wards and periods of unused capacity. An empty bed in one specialty is not automatically usable for every patient. The annual percentage is not a tool for choosing a hospital and does not show today’s availability.
These checks cannot establish that all hospitals are crowded or that all are empty. They establish a more specific requirement for any public comparison: name the duration, bed category and calculation basis before turning the percentage into a conclusion about capacity.
Before counting empty beds, establish what is being counted.
CNAS’s publication makes precisely this kind of scrutiny possible. A more explicit label would distinguish use against a contracting reference from actual occupancy across a calendar year. Publishing each indicator’s formula and a history of capacity changes would let readers reconstruct the comparison.
The remaining questions are concrete. Which formula produced each record? What period does the bed count describe? What explains the 134 cases that reproduce neither reference? The register makes those questions inspectable.
How we checked
The analysis uses the public CNAS file downloaded on 3 September 2026. It declares the year 2025 and an extract from the Ministry of Health/DRNC reporting template dated 22 May 2026. One row represents one hospital. The main test uses only acute care.
For each row with three positive values, we compare the published proportion with patient-days ÷ contracted beds ÷ 290, and separately with the same fraction divided by 365. An absolute difference of no more than 0.005 percentage points counts as a match. Missing and zero values are not replaced. All 35 ineligible records remain in the register.
The threshold is deliberately strict. Widening it to 0.01 percentage points gives 208 matches to 290 days; at 0.1 percentage points, there are 215. The headline uses the 198 strict matches. We do not calculate a national “true occupancy” rate or extend the formula to unexplained cases.
This is a documentary and arithmetic investigation. It includes no hospital visits, interviews or confirmation from CNAS of the formula applied to every row. The instrument shows data archived on the analysis date, not a live census.
- CNAS file containing all 372 hospital records
- CNAS map and displayed definitions
- Launch announcement, 17 July 2026
- Rules consolidated in 2025, Annex 23, p. 471
Sources checked 3 September 2026. Editorial review due by 3 December 2026.