Occupancy Rate

Occupancy Rate in Hospital Statistics

In hospital statistics the occupancy rate almost always means bed occupancy, calculated either from a daily census of occupied and available beds or from total bed-days divided by available beds times the days in the period. Health economists use it to judge how intensively beds, a large fixed input, are used, to plan bed numbers, and as a warning sign, since persistently full hospitals are associated with worse outcomes. It can also be derived from admissions and average length of stay, and national counting rules differ. The general measure, the 85% and 90% thresholds and the costing of spare capacity are on the capacity utilisation page.

The daily census and the bed-day ratio

Occupancy is built from a daily count. Each day a census records how many beds were available and how many were occupied, and the rate divides the two totals.

$$OR = \frac{\sum_{d=1}^{n} O_d}{\sum_{d=1}^{n} A_d} \times 100 = \frac{\bar{O}}{\bar{A}} \times 100$$

where $OR$ is the occupancy rate in per cent, $n$ is the number of days in the period, $O_d$ is the number of beds occupied at the census on day $d$, $A_d$ is the number of beds available on day $d$, and $\bar{O}$ and $\bar{A}$ are the average daily numbers of occupied and available beds. Public Health Scotland publishes the rate in the second form, average occupied beds over average available staffed beds. The OECD collects total bed-days in the year divided by the number of available beds multiplied by 365, which equals the first form only when the bed count is constant over the year.

The census time and the meaning of "occupied" are choices. Under NHS England's KH03 rules, which the Nuffield Trust summarises, a bed on an overnight ward is occupied if a patient is in it at midnight and a day-only bed if at least one day case used it; the quarterly rate divides the average daily number of occupied beds by the average daily number available. Germany counts all inpatients at midnight, with the day of admission counted as one bed-day so that day cases are normally included, and Japan bases its rate on inpatients at midnight in general and infectious disease beds. American Hospital Association data for the United States divide the average daily census by beds. Public Health Scotland also counts a bed reserved for a patient on pass as occupied. A midnight count misses patients who arrive and leave between two midnights, and other counts use other hours; de Bruin and colleagues compared their model with occupied beds counted at 08.00.

Which beds count as available

The denominator is where most differences arise. A bed without staff, or a trolley in a corridor, may or may not be counted, and every exclusion raises the rate.

Public Health Scotland defines an available staffed bed as one resourced for inpatient or day case care, and counts allocated beds plus borrowed beds, minus lent beds, plus temporary beds. Germany excludes cots for healthy infants, recovery trolleys, emergency stretchers and palliative care beds from the beds used for occupancy. The Netherlands subtracts beds for day care from curative beds, and from 2007 counts beds approved by the Dutch Health Authority instead of actual beds. Norway counts beds at the end of the year. Mismatches also occur. The OECD note on Belgium's psychiatric curative care rate states that it is underestimated because bed-days exclude psychiatric hospitals while beds include them.

Comparisons need care. The OECD's note for the United Kingdom states that its figure is based on the fourth quarter of each financial year, January to March, so the UK's annual entry is a winter quarter. Public Health Scotland reports provisional occupancy of 88.5% for acute specialties in 2025/26, down from 88.9% in 2024/25, a figure that ranks against other systems only where the definitions match.

Occupancy from admissions and length of stay

When admissions and stays are stable, occupancy can be computed without a census. De Bruin and colleagues define ward occupancy through Little's formula, in which the average number of occupied beds equals admissions per unit of time multiplied by the average length of stay.

$$OR = \frac{\lambda \bar{L}}{B} \times 100$$

where $\lambda$ is admissions per day, $\bar{L}$ is the average length of stay in days and $B$ is the number of available beds. Bosque-Mercader and Siciliani derive the same identity in a steady state, writing occupancy as length of stay multiplied by daily discharges per available bed.

Ireland's OECD note for curative care in public hospitals in 2023 shows the identity in annual figures. It reports 655,675 discharges and 4,059,959 bed-days, so the average stay is 4,059,959 / 655,675 = 6.19 days, matching the reported 6.2. With occupancy of 93.4%, the implied number of available beds is 4,059,959 / (365 × 0.934) = 11,909, or about 11,900 once rounding in the published rate is allowed for.

The identity shows that occupancy has three drivers, and a rise can come from more patients, longer stays or fewer beds; the rate alone does not say which. The average length of stay page gives the annual form of the identity, and the bed turnover page shows how occupancy and length of stay together fix the turnover rate and the empty-bed interval, including the charts that read all three at once.

High occupancy and patient outcomes

High occupancy is often read as efficiency, since fixed beds serve more patients, but it can also lower quality. Bosque-Mercader and Siciliani list the channels, which include patients placed on unsuitable wards, delayed admissions and cancelled operations, premature discharge to free beds, and heavier staff workloads that raise the risk of adverse events and infection.

Three studies estimate the size of these associations. Madsen, Ladelund and Linneberg analysed all 2.65 million admissions to Danish departments of medicine from 1995 to 2012 and found that high occupancy was associated with a 9% increase in in-hospital and thirty-day mortality compared with low occupancy. Friebel and colleagues studied 136 non-specialist English trusts between April 2014 and February 2016, when average occupancy was 90.4%, and found that a 1% increase in occupancy was associated with a 0.49% rise in the share of overnight patients discharged and a 0.011% increase in 30-day readmission, with stronger associations above 95%. Bosque-Mercader and Siciliani, using NHS England's overnight general and acute occupancy for 2010/11 to 2017/18, found that a 5 percentage point increase, about one standard deviation, was associated with 1.1% higher overall mortality, 3.1% higher surgical mortality and 0.5% to 0.9% lower patient-reported health gain after hip and knee replacement. They found no association with condition-specific mortality or emergency readmission; length of stay explained 12% to 25% of most of the associations, which mostly reflected differences between hospitals, not changes within them.

These are observational associations, not the effect of adding or removing beds. For economic evaluation the more direct point is that the value of bed-days released by an intervention depends on occupancy. In a full hospital the freed beds are refilled, so a released bed-day is worth the benefit of the care other patients now receive, the opportunity cost the bed-day carried; where beds stand empty it saves little beyond variable cost.

Worked example with one ward counted three ways

All figures are illustrative. A medical ward has 24 beds over a 30-day month, but 4 beds are closed for 10 days. The midnight census sums to 612 occupied bed-days, and the ward records 136 admissions with an average stay of 4.5 days.

1. Bed-day ratio. Physical bed-days are 24 × 30 = 720, and the closure removes 4 × 10 = 40, leaving 680 available bed-days. Occupancy is 612 / 680 = 0.90, or 90%. Measured against the 720 physical bed-days, the same patients give 612 / 720 = 0.85, or 85%, so the closure alone adds 5 percentage points.

2. Admissions and length of stay. Admissions average 136 / 30 = 4.53 a day, and 136 × 4.5 = 612 bed-days with every stay falling inside the month, an average of 20.4 occupied beds. Average available beds are 680 / 30 = 22.67, so occupancy is 20.4 / 22.67 = 0.90 again. The two routes agree because the census and the stays describe the same patients.

3. Reading the rate with length of stay. The ward had 68 empty bed-days (680 available less 612 occupied), or 68 / 136 = 0.5 days (12 hours) of empty time per admission, the turnover interval on the bed turnover page. If the average stay fell to 4.0 days with the same admissions, occupied bed-days would be 136 × 4 = 544 and occupancy 544 / 680 = 0.80, or 80%.

4. Interpretation. The ward's 90% is partly a product of closing beds, since the same patients give 85% against physical beds, and half a day less per stay would give 80%. Occupancy figures for two wards or hospitals are comparable only when their denominators match and their lengths of stay are read alongside.

Where occupancy figures mislead

The rate compresses a daily distribution into one number. Most misreadings come from forgetting what was counted and over which beds.

• A hospital average hides its wards. In de Bruin and colleagues' data, ward rates in 2006 ran from 44% on the paediatric intensive care unit to 85% on an internal medicine ward, and a small specialist unit can be short of beds while the hospital average looks comfortable.
• The same rate means different pressure on small and large units. A rate shows how full beds were on average, not how often a patient found none free, and de Bruin and colleagues showed that larger units can run fuller than smaller ones for the same share of refused admissions. The capacity planning page works this through with Erlang queueing models and explains why a single occupancy target does not fit every unit.
• A rate above 100% signals a counting mismatch. The Nuffield Trust notes that occupancy under KH03 rules cannot exceed 100%. A higher figure can mean patients in beds outside the counted denominator, as when wards temporarily open extra beds beyond their operational count, which de Bruin and colleagues observed on busy days. They also note that the hospitalised-days definition used in most Dutch hospitals allows rates above 100%.
• State occupancy is a different idea. In a Markov model, state occupancy is the time a cohort spends in a health state and has nothing to do with beds.

Sources

• Bosque-Mercader L, Siciliani L. The association between bed occupancy rates and hospital quality in the English National Health Service. The European Journal of Health Economics. 2023;24(2):209-236. https://doi.org/10.1007/s10198-022-01464-8
• de Bruin AM, Bekker R, van Zanten L, Koole GM. Dimensioning hospital wards using the Erlang loss model. Annals of Operations Research. 2010;178:23-43. https://doi.org/10.1007/s10479-009-0647-8
• Friebel R, Fisher R, Deeny SR, Gardner T, Molloy A, Steventon A. The implications of high bed occupancy rates on readmission rates in England: a longitudinal study. Health Policy. 2019;123(8):765-772. https://doi.org/10.1016/j.healthpol.2019.06.006
• Madsen F, Ladelund S, Linneberg A. High levels of bed occupancy associated with increased inpatient and thirty-day hospital mortality in Denmark. Health Affairs. 2014;33(7):1236-1244. https://doi.org/10.1377/hlthaff.2013.1303
• Nuffield Trust. Hospital bed occupancy. QualityWatch indicator, using NHS England Bed Availability and Occupancy (KH03) data. Accessed 3 October 2026. https://www.nuffieldtrust.org.uk/resource/hospital-bed-occupancy
• OECD. OECD Health Statistics 2025: Definitions, Sources and Methods. Hospital aggregates: curative (acute) care, with country notes. Paris: OECD; July 2025. https://stats.oecd.org/wbos/fileview2.aspx?IDFile=efc0d5e1-1350-4d26-a1ff-7d17533dede5
• Public Health Scotland. Acute hospital activity and NHS beds information (annual): year ending 31 March 2026. Edinburgh: Public Health Scotland; 29 September 2026. https://publichealthscotland.scot/publications/acute-hospital-activity-and-nhs-beds-information-annual/acute-hospital-activity-and-nhs-beds-information-annual-year-ending-31-march-2026/