Specificity of Healthcare Providers' Responses to Pay-for-Performance Incentives in the United Kingdom

The United Kingdom's Quality and Outcomes Framework (QOF), an ambitious pay-for-performance program implemented in 2004, changed a wide variety of incentives for primary care practices. In this paper, I use electronic medical records data from 357 practices to analyze the specificity with which practices responded to the new incentives, focusing on the following discrete thresholds around which marginal incentives are discontinuous: (1) test score thresholds that define success or failure on a quality indicator, (2) performance ceilings above which practices do not receive additional payment, and (3) end-of-fiscal-year effects introduced by annual reporting of results.

I find discrete differences in behavior around each of these thresholds, but find that practices were not entirely sophisticated in their responses. Specifically, test score thresholds (such as a practice being rewarded based on the percent of diabetic patients with cholestrol below a certain value) introduce an added incentive to retest patients above the threshold and a disincentive to retest patients below the threshold. I find that the QOF increased retesting above the threshold, but did not decrese retesting below the threshold. Furthermore, this increase was more marked among practices that were below the performance ceilings above which they would not receive additional payment.

I also analyze patterns of practices claiming exceptions on patients (removing the patient from the denominator of the performance measures). I find that practices operating below the performance ceilings were more likely to claim exceptions for their diabetic patients. Furthermore, there was strong seasonality in exception reporting, with much higher rates occuring in the last three months of the fiscal year. Overall measured performance was also markedly higher at the end of the fiscal year (when payments are calculated) compared to other times in the fiscal year.

This paper shows changes in provider behavior that are difficult to attribute to anything other than pay-for-performance (along with public reporting of results), but some unintended consequences that might have been predicted failed to materialize.