Top-Down Costing

Top-Down Costing

Top-down costing begins with an aggregate pool of recorded expenditure or resource cost and allocates it to services, activities or patients. It is used when complete service-level accounts are available but the resources used in each individual encounter are not measured directly. The resulting unit cost depends on the cost boundary, the allocation rules and the number and type of output units; it is not automatically the additional cost of treating one more patient.

What is allocated, and to what?

A costing study first identifies its cost object: for example, an outpatient visit, inpatient bed-day, procedure, completed episode or treated patient. It then determines which costs belong in the analysis, the time period and the perspective. The cost object matters because dividing the same total by visits, episodes or bed-days yields different units that answer different questions.

Directly attributable costs can be assigned to a service where the accounting data allow it. Shared costs such as facilities, management, information technology and support staff need an allocation driver. Common drivers include staff time, occupied area, equipment use, workload measures or service volume; the driver should plausibly reflect why the shared resource is used. A cost centre is an organizational unit whose costs are collected before allocation to final services. In a step-down allocation, support-centre costs are distributed in a defined sequence to other centres; reciprocal use between support centres may require a different allocation method or a disclosed approximation.

The basic average-cost relationship is

$$\text{Unit cost of service }j=\frac{D_j+\sum_k a_{kj}O_k}{N_j},\qquad \sum_j a_{kj}=1,$$

where $D_j$ is directly assigned cost for service $j$, $O_k$ is shared cost pool $k$, $a_{kj}$ is the share of pool $k$ allocated to service $j$, and $N_j$ is the number of comparable service units produced in the period. The summation condition applies when the entire pool $k$ is within the stated scope and allocated once across the included services. If some cost is deliberately excluded or retained centrally, disclose and reconcile that amount rather than pretending the allocated shares sum to the original organizational total.

Worked example: two outpatient services

Suppose a clinic has £900,000 of costs directly attributable to service A and £600,000 to service B in one year, plus £900,000 of shared support cost. Staff records indicate that A uses 6,000 hours of the shared support activity and B uses 4,000 hours. A delivers 12,000 comparable visits and B delivers 4,000. These are illustrative accounting figures; the hours are an allocation driver, not additional expenditure to add to the £900,000 pool.

The allocation reconciles exactly: £1,440,000 plus £960,000 equals the £2,400,000 cost within scope, and multiplying each service's cost per visit by its own visits returns its allocated total. Service B has a higher average unit cost in this example because its cost and volume differ; the table alone does not establish that B is inefficient or that its next visit would cost £240.

Now suppose the shared pool is allocated by visit count instead of recorded support hours. A would receive $12{,}000/16{,}000=75\%$ of it (£675,000), and B 25% (£225,000). The alternative unit costs become £131.25 for A and £206.25 for B; both allocations still reconcile to £2,400,000. The difference is caused solely by the assumed shared-cost driver, so the driver should be justified and, where important, varied in sensitivity analysis. Neither allocation is intrinsically more accurate without evidence about actual support use.

From accounts to an auditable unit cost

Good top-down costing preserves a trail from source records to the final unit. The analyst should specify the accounting period, decide whether cash expenditure is an adequate proxy for economic resource cost, and adjust for items such as capital use, donated resources or one-off charges when the decision perspective requires them. Then each shared pool needs an explicit driver and a rule for handling missing or mixed activity data.

1. Define the perspective and output. State the decision, included resources, provider or payer boundary, patient group, service unit and price year.
2. Reconcile the source totals. Match the included ledger or budget categories to the reported total, and document exclusions, transfers and any valuation adjustments.
3. Assign traceable costs. Place costs directly into final services when their use is observable and attribution is credible.
4. Allocate shared pools. Choose and record drivers, allocate each pool once, and disclose the method for support centres that serve one another.
5. Count comparable units. Align activity counts with the same period, service definition and eligible population as the cost numerator; account for case mix where needed.
6. Calculate and test. Divide the fully allocated service cost by its corresponding output count, check reconciliation, and assess sensitivity to the allocation and volume assumptions.

An activity count of zero makes an average unit cost undefined rather than zero. If fixed capacity remains while throughput falls, the average cost per unit can rise even if total spending is unchanged. Analysts should report capacity and utilization where they materially affect interpretation.

What top-down costs can and cannot answer

Top-down estimates are useful for describing full average provider costs, valuing common services in an economic model and identifying broad cost differences when patient-level resource measurement is infeasible. They may be sufficient for a decision that uses average service prices or nationally representative unit costs, provided their included resources, setting and case mix fit the question. Aggregate data can also make overhead inclusion more systematic than an incomplete list of observed consumables.

But an average allocated cost is not necessarily a marginal cost, which is the change in total cost caused by one additional unit at the relevant scale. A service with spare capacity may incur little extra fixed cost for one more visit; sustained expansion may require an additional team or facility and cause a step increase. A reimbursement tariff or charge may differ from provider resource cost, and an accounting expense can differ from opportunity cost. The correct valuation for economic evaluation depends on perspective, decision and the applicable reference case.

The top-down/bottom-up distinction describes where measurement starts: totals and allocation versus identifying and valuing resources used in individual activities. Gross costing versus microcosting describes the level of resource detail. These dimensions can overlap without being identical: activity-based allocation can bring an aggregate pool down to detailed activities, while a patient-level study may use a gross unit price for an inpatient day. Report both the measurement granularity and valuation/allocation method instead of calling every top-down estimate “gross costing.”

Case mix, comparability and double counting

One average price can hide variation in severity, length of stay, staff mix, location and technology. If the new technology changes those components, an old average for a broad service group may fail to represent the incremental resource use. Stratification by clinically meaningful service group, patient type or activity can improve relevance if the data support it; arbitrary subdivision can produce unstable estimates.

Check whether a unit cost already includes overhead, diagnostics, pharmaceuticals or capital. Adding those components again to a fully allocated amount double counts them; omitting them from a narrow estimate can understate total cost. Also distinguish a provider's full cost from the narrower cost borne by the payer or by patients. When transferring a published unit cost, inspect its source method, currency, price year, setting, included components and denominator before applying inflation or local adjustments.

For interventions that change volume, the denominator is particularly sensitive: $\text{average unit cost}=\text{total included cost}/\text{output}$ can change even if no input price changes. Avoid multiplying an average full unit cost by a large projected expansion without considering capacity, fixed versus variable resources and whether the historical unit is the same as the future service.

Reporting and quality checks

Report the source accounts and period, perspective, cost object, included and excluded categories, valuation adjustments, cost-centre structure, each allocation pool and driver, service counts, currency and price year. Present numerator and denominator alongside each unit estimate so it can be reproduced. If local costing assumptions are uncertain, show how key unit costs and the decision result respond to plausible alternatives.

• Arithmetic: Allocated shares should match their intended pool, and final service totals plus explicit exclusions should reconcile to source costs.
• Population and period: Activity counts must cover the same services and dates as their cost numerator, with returns, partial episodes and zero-activity centres handled explicitly.
• Units: Cost per visit cannot be substituted for cost per patient or cost per bed-day without a justified conversion.
• Relevance: The case mix, technology, local prices and resource bundle should fit the application; otherwise adapt or supplement the estimate.
• Decision scale: Distinguish historical average cost from the incremental and capacity costs that matter for the proposed change.

Sources and further reading

The WHO-CHOICE programme-costing material discusses service and overhead components in health-sector costing. The WHO health-service price-setting chapter describes average, top-down approaches and allocation to cost centres. CHEERS 2022 explanation and elaboration distinguishes costing granularity from valuation methods; the NICE economic evaluation manual supplies reference-case context for relevant costs and unit prices. The clinic figures are an original teaching example, not observed provider data.