摘要：This paper studies hospitals' upcoding and its implications for the cost-effectiveness of diagnosis-based prospective payment system (PPS). We focus on a payment reform in the public health insurance for urban citizens in city S in China. The reform introduced a system with diagnosis-specific scores, and one year after the introduction, it changed the scores for 55\% diagnoses. Using the individual-level inpatient discharge records, we compare the variations in diagnosis scores before and after the reform for individuals in city S versus that for individuals in a neighborhood city N. Results suggest that hospitals tend to upcode patients' diagnosis score by 5.8\% -8\% under diagnosis-based PPS relative to fee-for-service, and patients are less likely to be coded to diagnoses subject to a score decrease after the score change. This leads to excess expenditure to health care.