PAY-FOR-PERFORMANCE FOR HEALTH SERVICE PROVIDERS Effectiveness, Design, Context, and Implementation

Countries are increasingly implementing pay for performance (P4P) as a way to improve health services. The evidence base is conflicting and difficult to interpret. It is necessary to more systematically explore evaluations of P4P schemes in order to synthesize more useful evidence to inform the use of P4P schemes in health care.
This thesis starts with a literature review, which shows that the results of evaluations of P4P schemes are heterogeneous, which may possibly be explained by differences in programme design, context, implementation, and evaluation study design. I sought to find ways to better analyse and make sense of these evaluations using two approaches. A quantitative approach was used to systematically explore the heterogeneity. I developed and tested a theoretical typology to categorise P4P schemes by their design features. This typology considers who receives the incentive, type of incentive, size of incentive, and perceived risk of not earning the incentive. I then used the typology to quantitatively explore the influence of P4P design features and evaluation designs on it effectiveness using meta-regression and multilevel logistic regression analyses. I also undertook a formative evaluation of a pilot P4P scheme in Nigeria (a case study). This used semi-structured in-depth interviews with 36 purposively sampled health workers to explore how contextual and implementation factors (e.g. delay in incentive payment) influenced the impact of the scheme.
This research presents three notable and novel contributions to knowledge about P4P in healthcare. First a useful typology was developed, which can be used to help categorize, think about, structure and report P4P schemes in a standardized and theoretically informed way. Second, I show that P4P schemes with design features such as payment to groups, large incentive size (>5% of salary or usual budget), and low perceived risk of not earning the incentive are more likely to be effective compared to schemes characterized by payment to individuals, small incentives, and high perceived risk of not earning the incentive. In addition, I demonstrate that P4P evaluations without adequate controls over-estimate the effectiveness of P4P. Third, I show that contextual factors such as incentive payment delays, poor health worker understanding of the P4P scheme, and poor infrastructure affect the effectiveness of the Nigerian P4P scheme and need to be addressed in its future development.