Capacity Planning

Capacity Planning

Capacity planning estimates the people, space, equipment and time needed to deliver a health service at a defined volume and standard of access. It aligns expected demand with usable capacity across a pathway, including variation and existing backlog. A service can have enough capacity on average yet still produce long waits when arrivals are uneven, resources are not available together or a bottleneck limits flow.

Demand and capacity have units and timing

Demand is the work requiring service in a specified period, such as new referrals plus clinically necessary returns. Capacity is the amount of that work the service can complete in the same period under a stated staffing pattern, skill mix, hours and case mix. A nominal count of rooms, beds or employees is not capacity if the corresponding staff, equipment or downstream service is unavailable.

Quantity	What it means	Common measurement error
New demand	Cases newly requiring a service in a period.	Counting only accepted referrals and hiding unmet need.
Required workload	New demand plus follow-ups, rebookings and tasks generated by care.	Counting a pathway's first contact but omitting later appointments.
Nominal capacity	Theoretical output if planned resources operate throughout their available time.	Treating rostered hours as clinical hours without leave, breaks or setup.
Usable capacity	Output possible with jointly available resources and realistic scheduling.	Multiplying room slots without checking staff or equipment.
Throughput	Cases actually completed in the period.	Equating completions with capacity when slots went unused or demand was lower.
Backlog	Cases already waiting at the period boundary under a defined queue.	Combining people of different priority, eligibility or referral status.

These quantities need a common service unit and time window. “Visits per week” and “clinician hours per month” cannot be compared until appointment duration and calendars are reconciled. Capacity to start treatment may not equal capacity to complete a diagnostic-to-treatment pathway.

How a queue forms

At a simple weekly level, let $B_t$ be the number waiting at the start of week $t$, $A_t$ the arrivals during that week and $C_t$ the maximum service completions available to that queue. If each case requires one slot, service is work-conserving and there are no priority, cancellation or abandonment complications, a useful accounting identity is

$$B_{t+1}=\max{0,B_t+A_t-C_t},\qquad D_t=\min{C_t,B_t+A_t},$$

where $D_t$ is the completed work. This is a queue-accounting illustration, not a waiting-time forecast for a complex hospital. If average arrivals persistently exceed average capacity, backlog grows; if capacity exceeds demand, the excess can clear a backlog only when the required resources are available for that work. Arrival, service-time and capacity variation can produce waits even when average capacity exceeds average demand.

Utilization is completed workload divided by available usable capacity for a specified period and resource. A service near 100% utilization has little room to absorb random arrivals, emergencies or service-time variation. A low aggregate utilization figure can coexist with long waits for a specialized subgroup whose relevant capacity is fully occupied. No universal “ideal utilization” applies across all services and safety requirements.

Worked example: the equipment bottleneck

Suppose a diagnostic service needs a room, qualified staff and a device for each appointment. Per week the rooms could support 360 appointments, the available staffed hours 240, and the device schedule 180, all after conversion to comparable appointment units. With all three resources required simultaneously and no other constraint, the device limits maximum usable capacity to 180 appointments per week. Assume average new demand is 150 appointments per week and there is an existing backlog of 60.

Weekly measure	Value	Interpretation
Room capacity	360 appointments	Rooms are not the immediate limit.
Staff capacity	240 appointments	Staff exceed device capacity in this setup.
Device capacity	180 appointments	The modeled bottleneck is the device.
Average new arrivals	150 appointments	Average demand uses 83.3% of bottleneck capacity.
Opening backlog	60 appointments	These are already waiting before the first week.

With exactly 150 new arrivals and 180 completed appointments each week, the simplified balance is 60 at the start, 30 after week 1 and zero after week 2. This deterministic two-week clearance is conditional on all assumed slots being usable and demand staying at 150. If instead 190 people arrive in an otherwise empty-backlog week, at least 10 remain waiting after the 180 device slots are filled; lower-demand weeks may clear them. The 83.3% average utilization does not imply that every arrival is served without delay.

Adding another room would not raise this service's 180-appointment bottleneck if the device and staff schedules stay as assumed. Additional device time could help only up to the next constraint, which is the 240-appointment staff capacity. A credible plan would also examine whether the real bottleneck shifts across days, whether urgent patients use protected slots and whether devices can operate during hours when staff are available.

Planning across a pathway

A patient may need triage, imaging, consultation, treatment and follow-up. Improving one step can move the queue to the next. Map the flow and count workload generated at each stage, including repeat tests and patients who move between organizations. For staffing, distinguish headcount from whole-time-equivalent clinical hours and account for supervision, training, leave, skill requirements and safe workload. For beds, occupancy and length of stay jointly shape flow; a bed is not operationally usable if nursing, cleaning, infection-control or discharge support is missing.

Planning can operate on several horizons. Short-term rosters handle daily and weekly variation, medium-term changes may require recruitment or extended hours, and longer-term investment may require buildings, equipment or redesigned pathways. Forecasts should reflect demographic change, disease burden, referral policy, treatment adoption and service productivity, with uncertainty rather than a single precise future count.

An arrival count can be suppressed by poor access: low observed utilization does not prove low need. Assess referrals turned away, delayed diagnoses, travel burden, equity by geography and priority group, and outcomes of waiting. A clinically safe service standard may call for reserved urgent capacity even if it lowers measured utilization.

Economic decisions and trade-offs

Extra capacity has costs, including staff, equipment, space and maintenance. Inadequate capacity can impose waiting, deterioration, avoidable admissions, lost health and staff pressure. A capacity plan compares feasible options such as extending hours, changing scheduling, improving flow, sharing equipment or purchasing additional capacity. Its economic evaluation should count both resource costs and health consequences over a relevant horizon, including implementation time.

Average cost per completed case can rise when spare capacity is held for peaks, but that reserve may provide valuable reliability and timely treatment. Marginal cost depends on whether the next case fits within existing staffed slots or requires a new shift or device; an average accounting cost does not answer that question. Budget impact concerns spending and timing for the provider or payer, while cost effectiveness asks whether additional health justifies the resources displaced. Throughput alone is not a health outcome.

Queueing or simulation models can incorporate variation, priority, no-shows, length of stay and interdependent stages. Their outputs depend on arrival patterns, service-time distributions, staffing rules and patient behavior. Validate them against observed waits, backlogs and utilization, and examine peak scenarios. A simple average-capacity spreadsheet remains useful for transparent first checks but should not be presented as a guarantee of access.

What to report and verify

State the population served, pathway boundary, service unit, planning horizon, capacity resource constraints, hours, priorities and access standard. Show observed and forecast arrivals by period, existing backlog, usable versus nominal capacity, completion counts and the assumptions used to handle variability. Report costs and health consequences from the stated perspective, and explain how the proposal affects constrained groups.

• Reconcile flow. Opening backlog plus arrivals minus completions and exits should match closing backlog under the actual queue definitions.
• Check joint resources. A completion requiring staff, room and device cannot exceed any simultaneously available bottleneck without changing workflow.
• Check case mix. A complex case may require more minutes or a different qualified team than an average appointment.
• Check variability. Use day-of-week, seasonal and urgent-demand patterns as well as averages when they matter to waits.
• Check displacement. An apparent improvement in one service may transfer waits or work to another stage or patient group.
• Check equity and safety. Measure access and outcomes for relevant populations rather than treating maximum utilization as the sole target.

Sources and further reading

The NHS England guide to aligning capacity with demand addresses demand variation, backlogs and usable service planning. Its demand-and-capacity models page highlights variability, rebooking and required capacity. The WHO workload indicators of staffing need provide a workforce-planning approach, and the Institute for Healthcare Improvement's hospital flow white paper discusses system bottlenecks and matching demand to capacity. The diagnostic example is an original teaching calculation.