COMPARISON OF BACTERIAL IDENTIFICATION SYSTEMS 259 Table V Causes of Identification Differences Between Conventional Methods and API a Cause of Conventional Identification Identification API Identification Differences Enterobacteriaceae E. gergoviae (4) b E. aerogenes URE(- ), CIT(- ), SORB(-) E. gergoviae (1) S. liquefaciens LDC ( - ), SORB( + ), GEL( + ) K. pneumoniae (1) Klebsiella species CIT(- ), URE(-) S. marcescens (2) S. liquefaciens Data base interpretation of neg. rhamnose c Nonj•rmentative Bacilli A. anitratus (1) P. paucimobilis N2( + ), GLU(- ), MEL(-) A. haemolyticus (1) A. anitratus Omission of A. haemolyticus from data base M. urethralis (1) Pasturella- Omission of TDA ( + ) Actinobacillus Moraxella from data base c P. maltophilia (1) P. cepacia ODC ( + ) P. pseudoalcaligenes (2) Pasturella- All test reactions A ct inobaci llus negative P. pseudoalcaligenes (1) Pseudomonas species Data base could not identify organism to species level P. putrefaciens (1) Pseudomonas species H2S (-) P. stutzeri (2) P. aeruginosa ADH ( + ) P. stutzeri (1) Achromobacter species Data base required P. stutzeri to be 100% URE ( - )c a First choice of the systems' spectrum of identifications. b Number in parenthesis indicates the number of strains misidentified. c Information deficiencies in the data base interpretation of biochemical test results contributed to or caused misidentification. bacter anitratus was identified as Cardiobacterium hominis, which illustrated an information deficiency in the system's data base. C. hominis was listed as 100% indole positive and A. anitratus as 100% indole negative. Although the microorganism was indole negative, it was still identified as C. hominis. Two strains of P. aeruginosa were misidentified because of positive urea tests. The MT data base did not allow for any urea positive strains of P. aeruginosa, even though positive urease production as high as 74% has been reported (6). One strain of P. maltophilia was not in the MT data base as a result of false positive anaerobic dextrose reactions and positive sucrose and xylose reactions. MT required P. maltophilia to be 100% sucrose negative and 91% xylose negative. However, reports of studies with conventional methods have found sucrose and xylose to be 92% and 56% positive (6) respectively. DISCUSSION Each of the methods utilized in this study had unique characteristics therefore, one would not expect identifications obtained with one system to correlate 100% with those

260 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS Table VI Causes of Identification Differences Between Conventional Methods and Flow a Cause of Conventional Identification Identification Flow Identification Differences Enterobacteriaceae E. cloacae (1) •' S. fonticola LDC (+) E. cloacae (1) S. liquefaciens LAC (-) E. gergoviae (1) K. pneumoniae ODC (-) E. gergoviae (1) Serratia fonticola SORB ( + ) E. coli (1) Serrati• odorifera ODC (-) Nonfer•entative Bacilli A. haemolyticus (1) A. anitratus Omission of A. haemolyticus from data base M. urethraIls (1) P. fluorescens/putida GLU( + ), XYL( + ), MALT( + ) P. maltophilia (3) P. paucimobilis DNA( - ), URE( - ), XYL( + ) P. pseudoalcaligenes (2) P. fluorescens/putida GLU (+), MANN (+), P. pseudoalcaligenes (1) A. xyloseoxidans XYL (+)c P. putida (1) P. aeruginosa ACET ( + ), 42øC ( + ) P. putida (1) P. picketti 42øC (-), LAC (-), MANN ( - ), MALT ( - ) P. putrefaciens (1) P. dimunta H2S (-) a •irst choice of the systems' spectrum of identifications. b Number in parenthesis indicates the number of strains misidentified. c Information deficiencies in the data base interpretation of biochemical test results contributed to or caused misidentification. of another. However, a parameter had to be established which would determine the accuracy of the five systems. The parameter decided upon was that of the systems' ability to correlate identifications with those obtained by conventional methods which were accepted as correct. The identifications obtained using API, FL, and MT were then compared to the conventional identification, evaluating the accuracy of identifi- cation, as well as that of individual biochemical tests. Inaccuracies in microbial identifications were encountered in all three systems. These included aberrant biochemical test results and inconsistencies in the information pro- vided by system data bases. The combination of these two problems exposed shortcom- ings in each system. The majority of the problematic identifications obtained with API were a result of false negative biochemical test reactions. This may have resulted from the system's apparent inability to detect positive reactions among the weakly sacchro- lytic NFB. The cause of this may be related to a lack of nutritive substances in either the biochemical substrates or in the inoculum vehicle. Difficulties were also encountered in the interpretation of the color changes in the decarboxylase tests. The manufacturer requires red or orange reactions to be considered positive at 24 h, and red only at 48 h. Frequently colors between orange and red were observed at 48 h, which resulted in. uncertainty in many of the final identifications. At best, the interpretation of the decarboxylase test color changes was very subjective. Among NFB, neither API, FL, or MT were able to differentiate A. haemolyticus from A. anitratus, as the former had not been incorporated into any of the respective data bases. API, however, is the only