Costing in economic evaluationBMJ 2000; 320 doi: https://doi.org/10.1136/bmj.320.7249.1597 (Published 10 June 2000) Cite this as: BMJ 2000;320:1597
This is another in an occasional series of notes on economics
These notes are edited by James Raftery ()
As argued in previous notes, the perspective of an economic evaluation—societal or confined to some organisation—affects the range of both outcomes and costs that should be included. The case has been made for taking a societal perspective,1 particularly for publicly funded programmes, but the range of resource effects that are potentially relevant is enormous. As with a stone dropped in a large pool of water the ripples in theory go on indefinitely, and some have spill-over effects. For practical purposes, however, one would want to measure only those occurring within a certain radius. The perspective adopted affects the “resource frame” of the study. The societal perspective implies taking a very large frame, while that of the public sector, or a particular organisation, implies a much reduced frame.
Three stages can be usefully distinguished in costing: identification, measurement, and valuation. Identification consists of listing the likely resource effects of the intervention as comprehensively as possible so that decisions can be made about the frame of the study—that is, which effects might reasonably be excluded. This decision depends on the perspective of the study. For example, a pharmaceutical company might restrict its perspective to the interests of the company. Even within a societal perspective, the decision about what to include in costing can be influenced by whether or not one wants to provide results which are comparable with other studies; for costing quality adjusted life years (QALYs) standardised approaches for “reference cases” have been proposed.2 Within a trial, restricting analysis to comparing the two or more arms may enable resource headings (non-healthcare charges, patients' and carers' time) common to each to be omitted. While this yields useful results, it restricts the comparability of the results to that service.
Measurement refers to the measuring of the resource changes included in the study. Typically these will be amounts of labour input or outputs (bed days, time in theatre, prescriptions) but may also include patients' or carers' time.
The final stage refers to valuation of these resource effects. If prices exist, and can be assumed to reflect costs, then these can be multiplied by the relevant units of service use to yield total costs, such as x bed days multiplied by £y per day. However, two problems complicate this simple picture: prices often do not exist for the relevant changes, and available prices may not reflect the societal value of resources. Economic theory suggests that prices will reflect resource values only under conditions of perfectly competitive markets—a situation which applies in part in the economy generally and hardly at all in health care. Economists have developed a range of methods for adjusting prices so that they better reflect resource use in studies with a societal perspective.1
Two strategies can be usefully distinguished in measuring and valuation: microcosting and gross costing. Microcosting refers to detailed analysis of the changes in resource use due to a particular intervention, similar to the time and motion studies. Such detailed, bottom up, collection of data on resource use may be necessary when changes are being made to existing services (adding an extra stage or test). With microcosting, valuation use is also likely to require customised work as prices are unlikely to be available. Although many analysts favour microcosting, it tends to be costly and runs the risk of being specific to particular contexts.
Gross or top down costing allocates a total budget to specific services such as hospital stays or doctors' visits according to rules.3 The simplicity of top down costing may be offset by a lack of sensitivity, which in turn depends on the type of routine data available. The choice between microcosting and gross costing depends on the needs of the analysis. Many studies use a mixture of the two, using microcosting for the direct costs of the intervention, and gross costing for other costs. Costs which are incurred long after the intervention (such as admissions in subsequent years) will, when discounted, be greatly reduced in value, which suggests gross costing for these.
The data sources available for costing depend on whether microcosting or gross costing has been adopted. Microcosting often relies on wage rates to value staff time. The use of national pay scales in the NHS provides readily available data. Drug prices are published in the British National Formulary; hospital inpatient costs at the level of healthcare resource groups are published annually 4 but with limitations 5; and the Department of Health funds an annual publication of the unit costs in community care.6
Finally, the costs and benefits of different approaches to costing have to be assessed. Though detailed and comprehensive microcosting is desirable in principle, practicality restricts it. For that reason alone, economists should arguably be involved in the early stages of study design.