OBJECTIVE:To examine the applicability of an Australian casemix classification system to the description of Chinese hospital activity. DESIGN:A total of 161,478 inpatient episodes from three Chengdu hospitals with demographic, diagnosis, procedure and billing data for the year 1998/1999, 1999/2000 and 2000/2001 were grouped using the Australian refined-diagnosis related groups (AR-DRGs) (version 4.0) grouper. MAIN OUTCOME MEASURES:Reduction in variance (R2) and coefficient of variation (CV). RESULTS:Untrimmed reduction in variance (R2) was 0.12 and 0.17 for length of stay (LOS) and cost respectively. After trimming, R2 values were 0.45 and 0.59 for length of stay and cost respectively. CONCLUSIONS:The Australian refined DRGs provide a good basis for developing a Chinese grouper.