Abstract： Objective To develop a method for rapid and accurate detection and identification of bacterial pathogens directly from body fluid specimens and to evaluate its application in clinical laboratory.Methods Bacteria DNA was extracted from 205 body fluid specimens with column-based kit,and the high variable V1 and V3 regions of bacterial 16S rRNA gene were amplified with broad-range primers.Amplicons were analyzed by pyrosequencing and the generated sequences were searched in the bacterial identification database.Traditional culture-biochemical method was also used for these specimens and the results were taken as the golden standard.SPSS 11.0 was used to calculate the sensitivity,specificity,false positive/negative rate,positive/negative predictive value and positive/negative likelihood rate of pyrosequencing method.Results The positive rate of bacteria culture was 39.5％ (81/205),among which 71 were infected with single bacterium,and 10 were infected with two species of bacteria.Compared with the culture identification results,pyrosequencing had a 100.0％ (71/71) concordance when applied to detect and identify bacterial pathogens from specimens with single specie bacterium infected.To specimens with two species bacteria infected,7 out of 10 specimens were in concordance with the culture identification results.Besides,pyrosequencing detected 10 positive specimens and identified bacterial pathogens infected in the 124 culture-negative specimens.Taken bacteria culture as the standard method,the sensitivity of pyrosequencing for identifying bacterial pathogen in body fluid was 100.0％,and with a specificity of 91.9％,the false positive rate was 8.1％,the false negative rate was 0.0％,the positive predictive value was 89.0％,the negative predictive value was 100.0％,and the positive and the negative likelihood rate were 12.4 and 0,respectively.Conclusion Pyrosequencing can be used to detect and identify bacterial pathogens directly from body fluid specimens with the advantages of rapidity,high sensitivity,high accuracy and high throughput.