• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jcmPermissionsJournals.ASM.orgJournalJCM ArticleJournal InfoAuthorsReviewers
J Clin Microbiol. Mar 2006; 44(3): 928–933.
PMCID: PMC1393076

Evaluation of the Phoenix 100 ID/AST System and NID Panel for Identification of Enterobacteriaceae, Vibrionaceae, and Commonly Isolated Nonenteric Gram-Negative Bacilli

Abstract

The Phoenix 100 ID/AST system (Becton Dickinson Co., Sparks, Md.) is an automated system for the identification and antimicrobial susceptibility testing of bacterial isolates. This system with its negative identification (NID) panel was evaluated for its accuracy in the identification of 507 isolates of the family Enterobacteriaceae, 57 other nonenteric gram-negative isolates that are commonly isolated in clinical microbiology laboratories, and 138 isolates of the family Vibrionaceae. All of the isolates had been characterized by using approximately 48 conventional tube biochemicals. Of the 507 isolates of the Enterobacteriaceae, 456 (89.9%) were correctly identified to the genus and species levels. The five isolates of Proteus penneri required an off-line indole test, as suggested by the system to differentiate them from Proteus vulgaris. The identifications of 20 (3.9%) isolates were correct to the genus level but incorrect at the species level. Two (0.4%) isolates were reported as “no identification.” Misidentifications to the genus and species levels occurred for 29 (5.7%) isolates of the Enterobacteriaceae. These incorrect identifications were spread over 14 different genera. The most common error was the misidentification of Salmonella species. The shortest time for a correct identification was 2 h 8 min. The longest time was 12 h 27 min, for the identification of a Serratia marcescens isolate. Of the 57 isolates of nonenteric gram-negative bacilli (Acinetobacter, Aeromonas, Burkholderia, Plesiomonas, Pseudomonas, and Stenotrophomonas spp.), 48 (84.2%) were correctly identified to the genus and species levels and 7 (12.3%) were correctly identified to the genus level but not to the species level. The average time for a correct identification was 5 h 11 min. Of the Vibrionaceae spp., 123 (89.1%) were correctly identified at the end of the initial incubation period, which averaged 4 h. Based on the findings of this study, the Phoenix 100 ID/AST system NID panel falls short of being an acceptable new method for the identification of the Enterobacteriaceae, Vibrionaceae, and gram-negative nonenteric isolates that are commonly encountered in many hospital microbiology laboratories.

The introduction of the Phoenix 100 ID/AST system (Becton Dickinson Co. [BD], Sparks, Md.) within the United States brings to four the number of automated identification systems on the market worldwide. The use of automated systems in clinical microbiology laboratories is widespread, and technologists rely heavily on their accuracy. Since bacterial identifications are linked to the algorithms for antimicrobial susceptibility testing in several of the machines, the accuracy of the identification affects the interpretation of the accompanying antimicrobial susceptibility tests. I evaluated the accuracy of the Phoenix 100 ID/AST system and the negative identification (NID) panel when they were used to identify members of the families Enterobacteriaceae and Vibrionaceae, members of the genera Aeromonas and Plesiomonas, and commonly isolated gram-negative nonenteric organisms.

MATERIALS AND METHODS

Identification system and software version.

The NID panels were processed in a Phoenix 100 ID/AST system. All panels were processed according to the manufacturer's directions.

Each NID panel contained 45 substrates plus two fluorescent positive control wells. The substrates used one of the following principles: enzymatic hydrolysis of the amide or glycosidic bond, which results in the release of a fluorescent coumarin or 4-methylumbelliferone derivative (13 substrates); resistance to an antimicrobial agent or utilization of a carbon source, which results in a reduction of the resazurin-based indicator (2 and 7 substrates, respectively); enzymatic hydrolysis of a colorless substrate, which releases a yellow end product (4 substrates); utilization of carbohydrates, which results in lower pHs and changes in the phenol red indicator (16 substrates); hydrolysis of ornithine or urea, which results in a change in the fluorescent indicator; or hydrolysis of esculin, which results in a black precipitate in the presence of the ferric ion.

Each panel must be inoculated within 2 h after its foil pouch is opened, and the panels must be loaded into the instrument within 30 min of inoculation. Only cotton-tipped swabs or wooden applicators are acceptable for preparation of the suspensions.

A suspension of each 24-h-old isolate was made in the Phoenix 100 ID/AST broth to match the turbidity of a 0.5 McFarland standard by using a CrystalSpec nephelometer (Becton Dickinson). The panel was inoculated, the inoculation port was sealed with a panel closure, and the panel was loaded into the instrument. The current database is version 3.34, which contains 60 genera, 155 species, and 5 CDC enteric groups.

Culture collection.

The 702 isolates of biochemically typical and atypical members of the families Enterobacteriaceae, Vibrionaceae, and Aeromonadaceae and commonly isolated gram-negative nonenteric organisms were taken from the stock culture collection of the Centers for Disease Control and Prevention (CDC) and had previously been characterized with 48 conventional biochemicals by standard methods (4, 6, 8). All isolates of Vibrio cholerae and Vibrio parahaemolyticus were serotyped for confirmation. Isolates were maintained in defibrinated sheep blood at −70°C. Upon removal from the freezer, the isolates were passed three times on tryptic soy agar with 5% sheep blood (TSA II; BD Biosciences Inc., Sparks, Md.) before inoculation into the NID panels. All incubations were at 35 ± 1°C, unless otherwise noted.

Eighteen isolates of biochemically typical and atypical Salmonella spp. were obtained from either clinical microbiology laboratories in the United States or the Salmonella reference laboratory at the CDC. These isolates had never been frozen and had been passed a minimal number of times since they were isolated from their respective patient source and before they were tested in the Phoenix system.

Additional tests.

The only additional biochemical test required by the Phoenix system for identification was the spot indole test to differentiate between Proteus penneri and P. vulgaris.

Definitions.

“Correct” means that the Phoenix system identification agreed with the reference biochemical identification at the genus and the species levels at the end of the incubation period. In this study, the incubation period ranged from 2 h 8 min to 12 h 20 min. “Correct to genus” means that the Phoenix system identified the organism to the correct genus but not to the species level, when that genus and species were included in the database. “No identification” means that the instrument could not identify the organism within the maximum allowable time of 12 h. “Error” means that the instrument misidentified the organism at a confidence value of ≥90% when that organism was contained within the database. A confidence level of 90% is the lower limit of acceptability for the Phoenix system. Any identification with a confidence value of <90% at 12 h is categorized as “no identification.” If an initial identification was in error, an additional passage on blood agar was made and the test was repeated in duplicate to eliminate the possibility of technical error. The best two of three answers were used for categorization of that isolate.

RESULTS

Table Table11 shows the results of testing of 507 isolates of the Enterobacteriaceae. At the end of the incubation period, 456 (89.9%) of the isolates were correctly identified, 20 isolates (3.9%) were correctly identified to the genus level only, and 29 isolates (5.7%) were incorrectly identified. Table Table22 expands on the errors in identification to the species level or totally incorrect identifications (error). These 29 errors were scattered over 14 genera and were not concentrated in any one particular genus, with the exception of the genus Salmonella. The identifications for 317 (62.5%) isolates that were correct were completed in 4 h or less.

TABLE 1.
Enteric isolates tested with NID card
TABLE 2.
Problems in identification

The database of the Phoenix NID groups together as “Salmonella species” isolates that are neither of the serovar Choleraesuis, Gallinarum, Paratyphi, Pullorum, nor of S. enterica Typhi, subsp. arizonae. Of the 16 stock Salmonella species cultures that have been frozen at −70°C for many years, 13 were correctly identified (81.3%) and 2 were misidentified (12.5%). In an effort to learn if the Salmonella misidentifications were the result of the isolates being frozen for many years, 18 fresh isolates were obtained from patients in community hospitals in 10 different states. These strains were passed only one time for a purity check before they were tested. Of the 18 fresh Salmonella isolates, 13 were correctly identified (72.2%) and 5 were misidentified (27.8%). What is particularly troublesome is that the seven misidentified isolates (stock or fresh) were called Escherichia coli with a confidence level exceeding 90%. In an effort to learn why these errors occurred, I worked with the manufacturer to test certain isolates in conventional biochemicals, with those results being compared to the results from Phoenix system testing. Raw data were given to BD, which selected five of the isolates for testing in 1% Andrade's galacturonate. Of those five isolates, four (80%) gave the same result with the conventional substrate that was seen with the Phoenix panel; however, they were still misidentified as E. coli.

Another disturbing set of results concerns the Shigella spp. The 10 isolates were not atypical in their biochemical results and should not have been difficult to identify, yet the instrument failed to identify 4 of them.

Table Table33 shows the results obtained from the testing of 57 isolates of seven nonenteric genera, of which 84.2% were correctly identified. The average time for a correct identification was 5 h 11 min, although the time was much shorter if the times for the Acinetobacter isolates are removed from the calculation. Table Table44 shows the results of testing of 138 isolates of eight different species of Vibrio and Photobacterium damselae. Of those isolates, 89.1% were correctly identified, although only 44.9% of the correct identifications were obtained in 4 h or less. Less than 3.6% of these isolates not in the family Enterobacteriaceae were misidentified. Table Table55 lists the identification errors for all the nonenteric isolates.

TABLE 3.
Nonenteric isolates tested with NID card
TABLE 4.
Vibrio spp. isolates tested with NID card
TABLE 5.
Problems with identifications of nonenteric bacteria

While the 702 isolates tested presented a true challenge to the instrument, they do not reflect what is more likely seen in the daily workload of the clinical laboratory. On the basis of input from local area hospitals, an assortment of strains that approximate the relative numbers and types of strains likely to be routinely isolated in a clinical microbiology laboratory was randomly extracted from the “challenge set.” This “weighted set” of 118 strains gave the results shown in Table Table6.6. Even with the weighted set of strains, it is recognized that some of the strains, if they were isolated in a clinical laboratory, would be subject to identification by using spot tests and would not be inoculated into a panel on an automated instrument.

TABLE 6.
Results of testing of a weighted set of strains

DISCUSSION

Many clinical microbiology laboratories require that their identification systems perform at an accuracy of 90% or better, and many require a 95% accuracy. The Phoenix system accurately identifies isolates of the family Enterobacteriaceae only 89% of the time. Even with a weighted set of organisms, the accuracy is 89%.

In the only other published study of the Phoenix system that used conventional biochemicals for a reference method, Colodner et al. reported 90.2% accuracy when identifying 51 isolates of Vibrio vulnificus biotype 3 as Vibrio vulnificus (2). Because the present study did not include any biogroup 3 isolates, it is not known how the present results compare to those of Colodner and coworkers (2).

Several evaluations that used other commercial identification products as the reference method have been performed. Endimiani et al. tested 136 nonfermenting gram-negative bacilli and reported 95.6% agreement between the Phoenix 100 ID/AST system and the ATB/ID32GN system (bioMérieux, Marcy l'Etoile, France) (5). All isolates of Pseudomonas aeruginosa and Stenotrophomonas maltophilia, perhaps the most commonly isolated nonfermenters in a hospital laboratory, were correctly identified. Stefaniuk et al. reported an accuracy rate of 92.5% compared to the results of testing with the API 20E system when they tested 120 isolates that represented only eight of the most commonly encountered species of Enterobacteriaceae (11). The same study showed an agreement of 96.3% compared to the results obtained with the API 20NE system for the identification of 54 isolates of P. aeruginosa, Acinetobacter baumannii, and S. maltophilia. When Donay et al. used the same two reference systems for comparison of the results to those obtained with the Phoenix system, the identifications of 130 isolates of the Enterobacteriaceae and 57 isolates of nonenteric organisms showed accuracy rates of 94.6% and 89.4%, respectively (3).

Brisse et al. tested 134 isolates of the Burkholderia cepacia complex from cystic fibrosis patients; they had been identified by five different molecular biology-based methods, and an accuracy rate of only 50% was reported (1). The rate of 85.7% in the present study is higher probably because none of the B. cepacia isolates in this study were from cystic fibrosis patients. These isolates are known to be more difficult to identify.

Schreckenberger et al. compared the Phoenix NID to both the Vitek Legacy and the Vitek 2 colorimetric systems (P. C. Schreckenberger, K. L. Ristow, and A. M. Krilcich, Abstr. 105th Gen. Meet. Am. Soc. Microbiol., abstr. C-193, 2005). Testing 288 isolates of Enterobacteriaceae and 129 isolates of nonfermenters, they reported NID accuracy rates of 93.8% and 83.7%, respectively.

In a related study by Funke and Funke-Kissling, in which 309 isolates were inoculated directly from positive blood culture bottles into the Phoenix NID panels, 92.9% of the isolates were correctly identified to the genus and species (7). At this time, that is the only study that has addressed the concept of identification without prior isolation of the organism in pure culture.

The Phoenix 100 ID/AST NID panels were easy to use. Once the suspension was made, the panel was inoculated, and the panel closure was snapped into place, the panel was completely sealed, thereby preventing possible contamination to the technologist. If, however, the panel was jostled unnecessarily or dropped, the liquid in the esculin well was disturbed enough that the baseline reading was not valid and the test with the panel was aborted. This study encountered three instances of panel closures that were mismolded during production. Because there were no obvious flaws in the closures, they were used, only to be ripped from the panel during the first rotation of the carousel. This action did not damage or jam the machine, but the test with the panel had to be aborted and set up again. BD is currently preparing to release an alternate inoculum procedure that uses an inoculum density equal to a 0.25 McFarland standard. This workflow has been validated for use with the current NID panels and allows the instrument to read panels inoculated either way simultaneously.

The Phoenix instrument requires a bench that is able to support 500 pounds and that has at least 6 linear feet of space. The machine accommodates 99 panels, with one slot allocated for the permanently installed thermometer. No internal cleaning or maintenance of the machine is required. The Phoenix system validates itself on every cycle.

All of the consumables that are required can be stored at room temperature. Panels are available in either “identification-only” formats or as combination panels with identification and antimicrobial susceptibility testing capabilities. The panels are read every 20 min; and calculations are made after the readings are taken at 2, 3, 4, 6, and 12 h. Panel testing ceases after 12 h 20 min.

The Institute of Medicine report To Err is Human: Building a Safer Health System proposed a comprehensive approach to reducing medical errors and improving patient safety (9). While laboratory errors were not addressed directly, one of the recommendations was that heath care organizations implement proven medication safety practices. Because of issues related to increased resistance to antimicrobial agents in certain genera, it is imperative that the identification of causative agents of infection be as accurate as possible, preferably in the shortest time possible, thus allowing appropriate antimicrobial therapy to be initiated.

In the 8th edition of the Manual of Clinical Microbiology (10), it is recommended that the accuracy of a system exceed 90% in its overall ability to identify common and uncommon bacteria normally seen in the hospital laboratory and that the system be able to identify commonly isolated organisms with at least 95% accuracy compared with the accuracies of conventional methods.

With an overall accuracy of 88.9% for the identification of a challenge set of enteric and nonfermenter organisms and an accuracy of 89.8% for the identification of the weighted set of isolates, the Phoenix NID panel falls short of providing accurate identifications to satisfy these criteria.

REFERENCES

1. Brisse, S., S. Stefani, J. Verhoef, A. Van Belkum, P. Vandamme, and W. Goessens. 2002. Comparative evaluation of the BD Phoenix and Vitek 2 automated instruments for identification of isolates of the Burkholderia cepacia complex. J. Clin. Microbiol. 40:1743-1748. [PMC free article] [PubMed]
2. Colodner, R., R. Raz, I. Meir, T. Lazarovich, L. Lerner, J. Kopelowitz, Y. Keness, W. Sakran, S. Ken-Dror, and N. Bisharat. 2004. Identification of the emerging pathogen Vibrio vulnificus biotype 3 by commercially available phenotypic methods. J. Clin. Microbiol. 42:4137-4140. [PMC free article] [PubMed]
3. Donay, J.-L., D. Mathieu, P. Fernandes, C. Prégermain, P. Bruel, A. Wargnier, I. Casin, F. X. Weill, P. H. Lagrange, and J. L. Hermann. 2004. Evaluation of the automated Phoenix system for potential routine use in the clinical microbiology laboratory. J. Clin. Microbiol. 42:1542-1546. [PMC free article] [PubMed]
4. Edwards, P. R., and W. H. Ewing. 1972. Identification of Enterobacteriaceae, 3rd ed. Burgess Publishing Co., Minneapolis, Minn.
5. Endimiani, A., F. Luzzaro, A. Tamborini, G. Lombardi, V. Elia, R. Belloni, and A. Toniolo. 2002. Identification and antimicrobial susceptibility testing of clinical isolates of nonfermenting gram-negative bacteria by the Phoenix automated microbiology system. Microbiologica 25:323-329. [PubMed]
6. Farmer, J. J., III, M. A. Asbury, F. W. Hickman, D. J. Brenner, and the Enterobacteriaceae Study Group. 1980. Enterobacter sakazakii: a new species of “Enterobacteriaceae” isolated from clinical specimens. Int. J. Syst. Bacteriol. 30:569-584.
7. Funke, G., and P. Funke-Kissling. 2004. Use of the BD Phoenix automated microbiology system for direct identification and susceptibility testing of gram-negative rods from positive blood cultures in a three-phase trial. J. Clin. Microbiol. 42:1466-1470. [PMC free article] [PubMed]
8. Hickman, F. W., and J. J. Farmer III. 1978. Salmonella typhi: identification, antibiograms, serology, and bacteriophage typing. Am. J. Med. Technol. 44:1149-1150. [PubMed]
9. Kohn, L. T., J. M. Corrigan, and M. S. Donaldson. 2000. To err is human: building a safer health system. National Academy Press, Washington, D.C.
10. O'Hara, C. M., M. P. Weinstein, and J. M. Miller. 2003. Manual and automated systems for detection and identification of microorganisms, p. 185-207. In P. R. Murray, E. J. Baron, M. A. Pfaller, J. H. Jorgensen, and R. H. Yolken (ed.), Manual of clinical microbiology, 8th ed. American Society for Microbiology, Washington, D.C.
11. Stefaniuk, E., A. Baraniak, M. Gniadkowski, and W. Hryniewicz. 2003. Evaluation of the BD Phoenix automated identification and susceptibility testing system in clinical microbiology laboratory practice. Eur. J. Clin. Microbiol. Infect. Dis. 22:479-485. [PubMed]

Articles from Journal of Clinical Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...