The effect of pay-for-performance on mortality in the United Kingdom

The Quality and Outcomes Framework (QOF) – the United Kingdom’s primary care pay-for-performance program – was introduced in 2004, linking up to 25 percent of family practitioners’ income to performance on 76 clinical quality indicators targeting specific disease areas as well as 70 indicators relating to patient experience and the organization of care.

In this study, we use population-level cause of death statistics between 1994 and 2009 to test whether the implementation of the QOF was associated with reduced mortality in its targeted disease areas compared to 20 other Organization for Economic Coordination and Development (OECD) countries.

Our study outcomes are age and sex-adjusted mortality rates per 100,000 citizens for coronary heart disease, cancer, diabetes, epilepsy, and hypertension. These are the disease areas that have been targeted consistently in the QOF since 2004 and for which adequate data is available for all countries. Following Nolte and McKee’s (2012) specifications on mortality amenable to health care, our calculation of mortality rates for each disease area includes deaths only at age 74 or younger, with lower thresholds imposed for diabetes (age<50) and certain cancers.

Our identification strategy used the synthetic control method described by Abadie and colleagues (2007). This method is designed to optimize the choice of comparison groups among multiple alternatives in difference-in-differences designs. For each disease area, we construct a “synthetic” United Kingdom as the weighted combination of one or more of comparison countries, where weights are derived to minimize the mean squared difference between the United Kingdom and comparison countries for pre-intervention factors. In our study, these pre-intervention factors are levels of the mortality rate per 100,000 citizens for each year prior to the start of the QOF in 2004 as well as gross domestic product per capita, employment for professions related to health and social services per capita, alcohol consumption per capita, and the Gini coefficient of income inequality. After the synthetic United Kingdom is constructed for each disease area, we performed a difference-in-differences analysis to test whether mortality rates declined more in the United Kingdom than in the synthetic United Kingdom in each of the first six years following the QOF. Statistical inference was based on permutation tests.

Our estimates indicate that the QOF was associated with a reduction in amenable deaths from coronary heart disease per 100,000 citizens by 0.31 in the 1^st year of the program (p=0.416), by 0.95 in the 2^nd year (p=0.375), by 1.27 in the 3^rd year (p=0.236), by 2.88 in the 4^th year (p=0.048), by 2.48 in the 5^th year (p=0.20), by 3.59 in the 6^th year (p=0.048), and by 11.48 over the first six years of the program (p=0.194). Our estimates suggest that by the 6^th year of the QOF, amenable deaths due to coronary heart disease fell by 11.7%. We found no evidence of reduced mortality for the other incentivized disease areas. More bonus revenue was at stake for coronary heart disease and other concomitant service changes targeting coronary heart disease may, in part, explain our findings.