Abstract: OBJECTIVE To compare the advantages, disadvantages and discriminatory power of three methods, namely pulse field gel electrophoresis (PFGE), multilocus sequence typing (MLST) and multilocus variable number tandem repeat sequence analysis (MLVA), in molecular epidemiological investigation of carbapenem-resistant Acinetobacter baumannii. METHODS A total of 212 strains of carbapenem-resistant A.baumannii were isolated from sputum specimens in a tertiary care hospital between 2011 and 2012, the molecular epidemiological survey was conducted for the strains by using PFGE, MLST and MLVA, and the discriminatory power and homogeneity were compared among the three kinds of methods with the use of Simpson's index of diversity and Wallace coefficient. RESULTS Of the 212 strains of carbapenem-resistant A.baumannii, 149 could be analyzed by PFGE, 200 could be analyzed by MLST, and 205 could be analyzed by MLVA.The results obtained by PFGE showed that type B was the dominant strain (102 strains, accounting for 68.46%).The results of MLST typing showed that ST195 was dominant (60 strains, accounting for 30%), followed by ST208 (43 strains, accounting for 21.5%) and ST643 (33 strains, accounting for 16.5%).The results of the MLVA typing indicated that type D was the predominant strain, (151 strains, accounting for 73.66%), followed by type N strains (8 strains, accounting for 3.9%), and the type K, M, T and W strains ranked the third place (4 strains of each type, accounting for 1.95%).The predominant strains were prevalent in the ICU.The Simpson correlation coefficients (D values) of PFGE, MLST and MLVA were 0.46, 0.76 and 0.94, respectively.The Wallace coefficients of MLVA and MLST were 0.76. CONCLUSION PFGE is an important tool for analysis of clonal time and spatial distribution in the region,the carbapenem-resistant A.baumannii strains that are prevalent in the hospital are mainly isolated from the ICU.MLST can conduct worldwide epidemiological studies on antibiotic-resistant strains and novel mutant strains.MLVA data can be used for exchange, integration, and analysis among different laboratories.The discriminatory power of MLVA is high, and the homogeneity of MLVA and MLST is high.