What would happen if individual football players were rewarded only according to the number of goals they scored? They would stop cooperating with other members of the team - possibly reducing the total number of goals scored - and they would forget defence, probably preventing the team from winning.

This is an important insight for the NHS as it tries to find a way of measuring the value of research and development. Valuing R&D is high on the agenda following the Culyer reforms and in the light of the forthcoming report of the Clarke committee. Publications are an often chosen indicator of research outputs - on their own they do not, however , capture those aspects of R&D that are most important.

The growing interest of the value of R&D is a product of policy directed to promoting a knowledge economy , with research seen as an investment.

1 These returns include not only economic gains, but also contributions to knowledge, enhanced research capacity , improved decision making, and health gains.

2,3 Second, the growing interest in valuing R&D reflects a desire for accountability in the use of tax funds. This is visible in the increasing use of performance indicators in a number of spheres, including the NHS s new performance framework.

4 What managers seek is a low cost measure giving accurate information about employee performance, which, if used to reward employees, also creates the right incentives.

To pick up the football analogy, take, for example, three Manchester United players: Andy Cole, Jaap Stam and goalkeeper Massimo Taibi. By the middle of October, Cole had scored a total of 89 goals for United, Stam had scored one and Taibi had not scored. The cost-effectiveness of each of these players could be calculated by comparing the number of goals they scored against their cost to the club, in much the same way that some commentators compare publications with research funding. Cole cost£6.25m, Stam cost£10.6m and Taibi cost£4.4m: is Cole the most cost-effective of these United players, since his cost per goal scored is the lowest?

This is obviously not the case. Goals fall foul of the first of the criteria listed: they do not measure all valuable activity. Players occupying different positions score goals at different rates and under differing circumstances, just as laboratory-based scientists and clinicians publish at different rates in different types of outlet.

A further criterion against which performance measures have to be assessed is their impact on incentives. Goals are desirable for football clubs, but ultimately the club does not want to maximise goals; it wants to win. We do not want goalies forward, trying to score goals: we want them in defence, stopping the other team s goals.

Like goals, information about publications meets the criteria of being verifiable and relatively low cost. This is very attractive in the context of NHS R&D, where hard pressed managers have to try to collect information from hard-pressed staff.

Alternative methods, such as expert review , user surveys and case studies, require substantive resources and/or greater input from researchers. They do, however , also provide information on other dimensions of the outcomes of R&D - particularly whether it has contributed to the efficient, equitable delivery of health gains.

The NHS is not interested in publications per se, but rather in R&D which promotes the efficacy, efficiency and equity of healthcare and is disseminated appropriately.

Publications can play a useful role in assessing the value of R&D, but they must be used with care if they are to measure what is valuable to the NHS: the impact on health.