Measuring outcomes in economic evaluations

Editor—The series of occasional notes on economics has undoubtedly helped clinicians to understand the key concepts and jargon used by economists. The note discussing the use of outcome measures in economic evaluation is, however, misleading.¹

Firstly, condition specific outcome measures and generic quality of life scales should not, in general, be used in cost effectiveness analysis.² The primary reason for this is that such scales do not have the requisite interval properties. The scores produced by the short form-36 questionnaire (SF-36), for example, are little more than transformed ordinal rankings.

Even if interval properties can be shown, the use of generic quality of life scales in cost effectiveness analysis is severely restricted by their production of a set of scores reflecting different domains of health. For example, if the SF-36 is used in an evaluation, it can produce conflicting cost effectiveness ratios with respect to its various dimension scores.² It is best to restrict the set of outcome measures appropriate for cost effectiveness analysis to those measured in natural units.³ Following such an approach would limit the direct use of health scales to cost consequences analyses—that is, the presentation of cost and outcome data in a disaggregated format.⁴

Secondly, the paper restricts its commentary on cost utility analysis to the use of utility based quality of life scales as the measure of outcome. Use of such scales is increasing and will become the dominant form of utility measurement in clinical research. However, health state valuation techniques, such as the standard gamble and the time trade off,⁵ may be used to produce study specific utility values. Such techniques are also the basis for the values awarded by the utility scales.

Thirdly, the economics note should point out that the use of willingness to pay and conjoint analysis is developmental, and their use in cost benefit analysis (as opposed to simple investigation of the patient’s preferences) is extremely rare. It should also be noted that both of these approaches are valuation techniques, as opposed to outcome measures. As such, they can be used only if the effects or characteristics of the interventions have already been measured.

We recognise that this series of short articles must simplify important issues to remain concise. As it stands, however, this economics note is likely to mislead readers regarding good practice and commonly accepted practice.