Effectiveness of Pay-for-Performance Incentive Designs on Diabetes Care

Taiwan's National Health Insurance program implemented a pay-for-performance (P4P) program based on process measures in 2001. In late 2006, the P4P was revised to also include achievement of outcome measures.

This study examined whether a change in P4P incentive design structure affected diabetes outcomes.

We used a longitudinal cohort study design using 2 population-based databases. Newly enrolled P4P patients with diabetes in 2002-2003 (phase 1) and 2007-2008 (phase 2) made up the study cohorts. Propensity score matching was used to match comparable cohorts in each phase. In total, 46,286 matched cohorts in phase 1 and 2 were analyzed. Process measures were defined as the provision of tests of glycosylated hemoglobin A1c (HbA1c), low-density lipoprotein cholesterol, and blood pressure, and outcome measures as changes in those values between baseline and last follow-up within 3 years. Patient-level generalized linear regression models were used and patient characteristics, physician characteristics, and health care facility characteristics were adjusted for.

Our results indicated that the process measures of HbA1c and low-density lipoprotein cholesterol tests did not differ significantly between the 2 phases. In addition, better improvements were noted in outcome measures for the phase 2 patients (ie, HbA1c level and lipid profiles), whereas nonincentivized intermediate measures (eg, blood pressure) showed no negative unintended consequences.

Quality of care tended to be better when both process and targeted outcome measures were combined as quality metrics in the P4P program in Taiwan.