Intended for healthcare professionals

Education And Debate

Marginal costs and benefits

BMJ 1996; 312 doi: (Published 06 January 1996) Cite this as: BMJ 1996;312:35
  1. David J Torgerson, research fellowa,
  2. Anne Spencer, PhD studentb
  1. a National Primary Care Research and Development Centre, Centre for Health Economics, University of York, York YO1 5DD
  2. b Department of Economics, University of York
  1. Correspondence to: Mr Torgerson.
  • Accepted 11 October 1995

Decision makers are interested in measuring the costs and benefits of various interventions, and sometimes they are presented with the average costs and benefits of alternative interventions and asked to compare these. Usually a newer intervention is being compared with an existing one, and the most appropriate comparison is not of average costs (and benefits) but of the extra—or marginal—costs (and benefits) of the new intervention. Reanalysis of the costs effectiveness ratio of biochemical screening of all women for Down's syndrome compared with age based screening shows that the marginal cost effectiveness of biochemical screening is pounds sterling47786, compared with an average cost effectiveness of pounds sterling37591. It may sometimes be difficult or costly to calculate marginal costs and benefits, but this should be done whenever possible.

Some economic evaluations compare alternative health care interventions only by assessing differences in the average costs and benefits and express their results as average cost effectiveness ratios. This ratio is calculated by dividing the total cost of a health care intervention by the total benefits of such intervention.

This approach can be misleading as it describes inaccurately the different costs and benefits occurring with alternative health care interventions. A better method is to compare the extra costs and benefits between different health care interventions in the form of a marginal cost effectiveness ratio. This is calculated by dividing the extra or incremental costs by the extra benefits of the intervention. This paper explains why marginal analysis is essential if decision makers are to use scarce NHS resources efficiently.

Example 1: Down's syndrome

Consider screening for Down's syndrome based on maternal age. In 20000 pregnant women 26 cases of Down's syndrome would occur, and eight of these (30%) would be in women aged 37 years or over, who comprise 5% of the pregnant population.1 If 500 women aged 37 years or older have an amniocentesis this detects four affected pregnancies with an amniocentesis cost of pounds sterling75000 and an abortion cost of pounds sterling4000, giving a total cost of pounds sterling79000. Hence the cost effectiveness ratio of maternal age screening is pounds sterling19750—that is, pounds sterling79000/4.

Changing from maternal age screening to biochemical screening requires marginal analysis to estimate the budgetary expansion needed to introduce biochemical screening and to describe its extra benefits. Only with such data can an efficient choice be made.

An evaluation of such a change was reported by Wald et al.1 They estimated that implementing biochemical screening would result in an average cost of about pounds sterling38000 for each affected birth avoided. In column 2 of table 1 the calculations are replicated for a population of 20000 pregnant women. The total costs of biochemical screening amount to pounds sterling413500. However, some of those costs (pounds sterling79000) would have been incurred in the absence of biochemical screening. Column 3 of table 1 shows that implementing biochemical screening requires pounds sterling290000 to fund extra test costs, pounds sterling37500 in extra amniocentesis costs, and pounds sterling7000 in extra abortion costs—thus, the screening budget requires expansion by pounds sterling334500, not pounds sterling413500. Hence, average cost analysis overstates the total cost of implementing biochemical screening by 24%.

Table 1

Marginal analysis of Down's syndrome screening

View this table:

In a similar fashion the benefits have been overstated by using average benefits. Thus, the extra benefit is seven avoided births, not 11, as four births would have been avoided anyway under the maternal age screening programme. Thus average benefit analysis overstates the benefits by 57%. Finally, the marginal cost effectiveness ratio is 27% greater than the average cost effectiveness ratio—that is, pounds sterling47786 not pounds sterling37591. A similar failure to undertake marginal analysis of biochemical screening was made by Piggot and colleagues.2

Example 2: hypertension

Failure to use marginal analysis occurred in an evaluation of shared care and nurse practitioner care for hypertension.3 Table 2 shows the total NHS costs for each scheme and the numbers of patients attending for review at the end of the evaluation (the outcome measure used). Shared care had more patients attending for review at a slightly higher average cost per review. Average analysis may lead purchasers to think that the extra benefits of shared care can be obtained for pounds sterling28.72 per patient. However, the key information required by purchasers is the additional cost per extra review achieved by shared care, not the average cost. From table 2 it can be seen that nurse practitioner care achieved a 75% review completion rate compared with 82% for shared care. The key benefit is the extra 7% of completed reviews achieved by shared care and comparing these with its extra costs. Hence the benefit of shared care is 18 extra patients reviewed for an extra cost of pounds sterling976. Therefore the marginal cost effectiveness of shared care is pounds sterling54 per extra reviewed patient, not pounds sterling28.72.

Table 2

Marginal analysis of shared care for hypertension

View this table:


In these two examples of economic evaluations we can see how the production and use of average cost and benefit data can mislead decision makers. Estimates of marginal costs and benefits are always preferable to average costs and benefits, and this has been advocated for several decades.4 5 Despite this, there are often large evaluation costs incurred by calculating marginal rather than average values,6 and in some cases this may justify using average costs. Indeed, health economists recognise the cost of collecting marginal cost information, and solutions, such as reduced datasets, have been proposed.6 In addition, sometimes coincidentally, average costs may equate to marginal costs. Nevertheless, the temptation to use average costs and benefits should be avoided whenever possible.

This note shows that if more care is taken in the economic analysis marginal values may often be derived with little or no extra research effort. Even when marginal costs and benefits are more difficult to estimate, the improved precision of the evaluation may justify the increased research effort. For example, if average costs had been used when evaluating an early discharge scheme for patients with hip fractures they would have overstated its financial benefit by 200%.7 For evaluations of competing interventions to produce valid results marginal costs and benefits should be used—not averages.

We thank Dr Brenda Leese and Professor Alan Maynard for helpful comments. The views expressed in this paper are not necessarily those of the Department of Health, who fund the national centre.


  • Funding Department of Health.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
View Abstract