The UK National Health Service has introduced a pay for performance scheme for its primary care general practices. For 65 indicators the scheme rewards practices for increasing the ratio of treated to eligible patients, up to an upper threshold. Eligible patients are the patients on the practice list who have been reported by the practice as having a particular condition (prevalent) and who have not been exception reported as not suitable for the indicator. We find that quality is lower in practices with more deprived patients, with more ethnic minority patients, and with larger patient lists. Behaviour is compatible with altruism: practices which were above the upper thresholds could have reduced the number of patients treated by 11.8% without reducing their revenue.

We use of the sharp discontinuity in reward at the upper threshold for indicators to test whether incentives affected reporting of prevalence and exceptions. Practices below the upper threshold in 2004/5 had an average exception rate of 8.55% in 2005/6 and we estimate that without the incentive to increase their exceptions they would have had an exception rate of 7.25 %. Practices which were above the upper threshold in 2004/5 had reported standardised prevalence in 2005/6 which was 3.5% greater, against a mean of 100, than if they had been below the upper threshold in 2004/5. (Joint work with Matt Sutton and Ada Ma)