Identifying and Measuring Spillover Effects in the Evaluation of Quality Incentive Schemes

UoM administered thesis: Phd

Abstract

Background: Theoretical evidence on multitasking warns against the use of high-powered incentives when healthcare providers may substitute effort away from non-targeted tasks. However, incentives may also have wider benefits when targeted and non-targeted tasks share commonalities in production. Evaluations of quality incentive schemes often overlook these potential spillover effects. I aimed to address this limitation by developing the existing empirical evidence on the spillover effects of quality incentive schemes. Methods: I first reviewed 37 empirical studies that had evaluated the spillover effects of 24 different quality incentive schemes. The findings from this review informed the design of three case studies. In each case study, I used a different empirical method to evaluate the between-patient spillover effect of a quality incentive scheme. The first two studies evaluated the Best Practice Tariff scheme, using (i) an interrupted time series model and (ii) an alternative model of effort substitution. The third study evaluated the spillover effect of the Advancing Quality scheme using a triple difference-in-differences model. Results: The interrupted time series and effort substitution models both identified positive spillover effects of the Best Practice Tariff scheme on the daycase rate of non-targeted patients in an incentivised surgical area. However, the interpretation of these models differed. The interrupted time series model identified an initial increase in the non-targeted daycase rate of 1.0 percentage points (95%CI, 0.5 to 1.6). In contrast, the effort substitution model identified that an increase in the targeted daycase rate of one percentage point was associated with an increase in the non-targeted daycase rate of 0.09 percentage points (95%CI, -0.02 to 0.19), relative to the underlying association in the before period. In addition, the triple difference-in-differences model also identified variations in the spillover effect between different non-targeted patients. Estimates from the model suggest that the Advancing Quality scheme increased the mortality rate of non-targeted patients that were treated by a non-exposed physician by 0.38 percentage points (95%CI, 0.06 to 0.70) but had an no harmful effect on patients treated by an exposed physician, in the long-term period. Discussion: Further research is needed on the mechanisms through which spillover effects occur. Evidence from the thesis suggests that patients benefitted from being treated by a consultant that was more exposed to patients with a targeted condition. Identifying when spillover effects may be positive has the potential to increase the benefits of future incentive schemes.

Details

Original languageEnglish
Awarding Institution
Supervisors/Advisors
Award date1 Aug 2020