OBJECTIVE： To compare the effects of 2 dispensing modes in Pharmacy intravenous admixture services （PIVAS）， and to promote the development of PIVAS. METHODS： Related data about drug dispensing within 92 days in the third quarter of 2015 were selected from PIVAS center branch （traditional dispensing mode， control group） and east branch （centralized dispensing mode， observation group） of our hospital， and then analyzed retrospectively. The management supporting measures were adopted for centralized dispensing mode. 2 dispensing modes were compared in respects of average dispensing time， average checking time， average preparation time， average total dispensing time， the amount of breakage， the amount of empty needle， the number of working errors， etc. RESULTS： Centralized dispensing mode was completed through adding software and hardware， arranging the staff order again， etc. on the basis of traditional one. Above 6 indexes of control group and observation group were 1.5 h and 0.7 h， 1.8 h and 1.1 h， 4.4 h and 3.4 h， 7.1 h and 5.9 h， 4.9 pieces and 2.8 pieces， 576 needles and 348 needles， respectively （P＜0.05）. The number of working errors were 0.58 times and 0.60 times in 2 groups （P＞0.05）. CONCLUSIONS： Centralized dispensing mode can shorten working time of each step， and reduce working strength， the amount of breakage and of empty needle； but it requires advanced software and hardware support. Traditional dispensing mode is more suitable for those hospitals with small-scale drug configuration and disperse drug types.