Published: 27/05/2004, Volume II4, No. 5907 Page 27

Is evidence-based medicine moving into a new phase - evidence-based choice? The notion that treatments should be tested for clinical effectiveness (evidencebased medicine) and cost effectiveness (economics-based medicine) is now, surely, uncontroversial. Ultimately, the logic of testing and the use of 'gold standard' evaluation methodologies such as randomised control trials has been overwhelming. The more recent acceptance that some cost-benefit test also needs to be applied before deciding to provide a treatment also suggests a recognition that opportunity cost cannot be ignored in such decisions.

The reason for testing is obvious: for a new intervention, there is uncertainty about the outcome. National Institute for Clinical Excellence guidance, based on technical appraisals of trial data, reduces uncertainty about outcomes - but not to zero. Aside from some of the known technical and statistical deficiencies of some trials (many drugs prescribed for children have not been tested on children, for example), for a patient in the real world there is an additional source of uncertainty about a treatment outcome: variations arising from differences in clinical practice, experience, skill, knowledge and a host of other differences relating, perhaps, to the quality of aftercare etc.

In the absence of the 'perfect' randomised control trial - in essence, large enough to deal comprehensively with (ie nullify) all these real-world variations - what are patients (and purchasers) meant to do if they want to reduce realworld uncertainty for themselves?

For both patients and purchasers, there is an economic process at the heart of current health policy: choice. But having the freedom to exercise choice (crudely, the ability to take your business elsewhere), while vital, is only the first step in reducing uncertainty. A crucial factor in linking the exercise of choice to improved outcomes and reducing realworld uncertainties is the presence of some signals which help patients and purchasers decide what to do. The usual market signal is price, but the new activity-based reimbursement system for healthcare providers introduced under payment by results deliberately fixes prices; they are planned to signal, well, nothing at all.

But even in traditional markets, prices are often only crude signals of the things consumers want to know about - there is not necessarily a straightforward and positive linear relationship between price and quality of the product, for example. This is partly because of the extraordinary lengths to which producers go to confuse such links in order to segment markets and extract maximum rent. Have you ever read, let alone understood, the tariff structure for your gas or electricity bills?

In the absence of meaningful price signals, what patients and purchasers require to make proper choices and hence, in process terms, to exert the Darwinian equivalent of environmental change on providers to encourage quality improvements, is information on what is important to them. And this, it turns out, is unrandomised, uncontrolled real-world information about 'reputation', 'quality' and health outcomes. All three are related, and none is even properly recorded in the NHS, let alone published.