There are a growing number of published studies that suggest that much health care is delivered inappropriately. There are calls for measures of appropriateness to be used by purchasers and others to regulate or influence the delivery of health care. This paper explores assumptions inherent in results generated by a leading measure of appropriateness and concludes that there are considerable uncertainties about the measure's meaning, the magnitude of bias that it contains, and the degree to which its application can be generalised. Some of these uncertainties could be resolved if the tacit assumptions inherent in the generation of the criteria could be made explicit. Existing measures of appropriateness are not yet sufficiently robust to be used with confidence to influence or control the delivery of health care. They may have a use as an aid rather than as a constraint in clinical decision making. A randomised controlled trial could resolve whether patients achieve better outcomes if their care is influenced by appropriateness criteria.

A leading article by Brook published in the BMJ recently identified appropriateness as “the next frontier” in the development of clinical practice.1 It argued that, firstly, there is too much information about medical practice for any doctor to assimilate all the information relevant to their practice. It is therefore impossible to “practise good medicine without additional help.” Secondly, for this (and other reasons) many patients receive care that is “inappropriate” (contributing to overuse of health care) and many others are not offered “appropriate” care (underuse of health care). Thirdly, the appropriateness of care can be measured, and, finally, the application of measures of appropriateness can reduce or eliminate both overuse and underuse of medical interventions.

These claims, if true, have huge implications for medical practice, given that some studies estimate that 20% to 60% of care …