Measures of health-related qualityof lifein animperfect world: a comment on Dowie

Professor Dowie has written an interesting and thought provoking paper on a long lasting debate in the literature on measuring health related quality of life. The debate between generic and condition specific measures (CSMs) has not progressed a great deal with time and he is right to question a purely psychometric approach that currently tends to focuses on effect sizes. He has also presented an interesting challenge to the compromise solution suggested by anumber of psychometricians to adopt both types of measure. The paper is predicated on the apparently useful distinction between knowledge- versus decision- validity, but this is something of a straw man argument. Advocates of CMSs are not interested in knowledge-validity but are concerned with different types of decisions. Furthermore, produ- cers of criteriafor judging the merits of mea sures (including ourselves (1,2)) do not presume it to be an absolute question of whether a measure is 'valid' but hope such lists assist researchers and other users to find the best measure for their purpose, such as informing clinical or resource allocation decisions. For patients and clinicians important informa- tion is provided by atria l through the use of CSMs in terms of the impact on the problem that they bring to the doctor - i.e. proximal concerns. Neither the patient nor the doctor are necessarily anticipating an impact on the more distal dimen- sions typically covered by generic measures. Professor Dowie's response to this narrow view is that there may be important consequences of the treatment that are not covered by the CSM and consequences for co-morbidities. This is the basis of the usual argument for using both types of measure in trials. Professor Dowie goes on to argue that a generic measure 'is intended to cover the full range of health outcomes' whereas the condition specific is by definition intended to cover an a rrower range. On the basis of this distinction of intention he argues that in most cases the GEN should be used alone, but he accepts that there may be circumstances where a CSM is preferable, but that it is never preferable to use both (for decision-making purposes). We would just like to make a few comments about this distinction of intention. CSMs typically focus on fewer domain of interest and hence may be able to ask more questions about each domain and thereby achieve a greater degree of refinement. For example, the EQ-5D has a large jump in scores between perfect health and any level of illness of 1.0-0.88 and has only two levels of imperfect health for each dimension. However, Professor Dowie seems sceptical as to the empirical evidence regarding the sensitivity of CSMs over generics. We agree there are cases that do not support this common claim in the psychometric literature (3), but equally there are many where this has been found including: EQ-5D in chronic pulmonary disease (4) and in cosmetic surgery (5), the SF-36 in urinary incontinence (6) and the example of urinary incontinence that he provides (7).