The paper demonstrates differential effects of a prospective payment system with declining per diem rates, dependent on the percentiles of length of stay. The analysis uses dynamic panel data estimates and a recent nationwide administrative database for major diagnostic categories in 1068 Japanese hospitals in 2006-2012 to show that average length of stay significantly increases for hospitals in percentiles 0-25 of the pre-reform length of stay and significantly decreases for hospitals in percentiles 51-100. The decline of the average length of stay is larger for hospitals in higher percentiles of the length of stay. Hospitals in percentiles 51-100 significantly increase their rate of nonemergency/unanticipated readmissions within 42 days after discharge. The decline in the length of total episode of treatment is smaller for hospitals in percentiles 0-25. The findings are robust in terms of the choice of a cohort of hospitals joining the reform. The paper discusses applicability of 'best practice' rate-setting to help improve the performance of hospitals in the lowest quartile of average length of stay. Copyright © 2015 John Wiley & Sons, Ltd.
Keywords: diagnosis-related groups; health care financing; health care quality; length of stay; prospective payment system; readmission rate.
Copyright © 2014 John Wiley & Sons, Ltd.