A pay-for-performance program in Taiwan improved care for some diabetes patients, but doctors may have excluded sicker ones.

Many countries have implemented pay-for-performance programs to improve the quality of care. The structure of these programs, however, can have perverse consequences beyond improving care for patients. To investigate this possibility, we studied the pattern of enrollment of patients with diabetes in the first five years of a pay-for-performance program in Taiwan's National Health Insurance Program from 2001 through 2005. Taiwan's program did sharply improve quality of care for enrolled patients, producing 100 percent or nearly 100 percent adherence to all process measures. But at the same time, only a minority of the nation's patients with diabetes were enrolled, because the program's design encouraged physicians not to enroll their most complicated patients. By "cherry-picking" the healthiest patients most likely to perform well on selected measures, physicians were able to game the system and potentially reap the rewards of higher pay-for-performance payments without actually improving the care of all of their diabetic patients. Our study provides a cautionary tale, emphasizing the importance of proper program design so that quality is improved on the broadest scale.