This study shows SSTs can be used to inoculate Phoenix ID broth to a 0.5 McFarland standard, as was also shown by Funke et al. for GNR . However, a 0.5 McFarland standard for GPC obtained by using SSTs was shown to consistently contain a lower inoculum than 1.5 × 108 CFU/ml. This may be due to the presence of blood culture components other than bacteria, since they could also contribute to the turbidity in the ID broth.
This study shows very good results for ID of Enterobacteriaceae. Only two errors occurred with ID in this group. One strain was not identified and one strain of E. coli was misidentified as S. choleraesuis. Results of ID for Pseudomonas species were less reliable. Both errors in this group were P. aeruginosa strains that were identified as P. fluorescens, a rare cause of bloodstream infections. These misidentifications did not lead to errors in interpretation of AST, but rare or unlikely results of ID should be dealt with carefully and be confirmed using additional tests. Other studies also showed that ID of non-fermenting GNR was less reliable than that of Enterobacteriaceae [18, 23]. This may be due to the lower growth rate of non-fermenters, which could result in weaker fluorescent biochemical reactions in the Phoenix ID panel. Errors in ID with the direct method could also be caused by traces of blood culture components in the ID broth. This however seems less likely, since with Enterobacteriaceae, errors in ID were rare.
Since the Phoenix system was not used for ID of GPC, ID by direct inoculation was not tested in this group. But since ID is required for interpretation of AST, in clinical practice, rapid AST will have to be combined with a rapid method of ID, such as PCR-based methods on whole blood, like LightCycler® SeptiFast Test MGRADE (Roche), VYOO Sepsis Test (SIRS-Lab), SepsiTest™ (Molzym), or MALDITOF-MS on positive blood cultures .
Some studies on direct methods for AST showed poor results for GPC [15, 16] or focus on GNR only due to unfavorable results for GPC . However, in this study, direct AST for Staphylococcus species and Enterococcus species showed good agreement with conventional methods, comparable to results of the standard method, but with fewer very major errors. Lupetti et al. , who tested the direct Phoenix method for GPC and compared their results with those of the Vitek 2, found an even higher agreement. They incubated a portion of the positive blood culture with saponin in order to harvest more bacteria from a positive blood culture through the release of intracellular bacteria.
Other studies that presented results of direct methods for AST of GPC showed variable results [13–16, 25, 26], which makes comparison difficult. But our results were comparable to those of the routine Phoenix method. Moreover, categorical agreement for most tested antibiotics in this study, including oxacillin and vancomycin, were well over 90% and the percentage of major and very major errors is low, meeting the standards proposed by Jorgensen et al. . Only erythromycin and trimethoprim-sulfamethoxazole showed lower agreements. The majority of errors for erythromycin were minor errors, but also some major errors occurred. Trimethoprim-sulfamethoxazole was the only antibiotic for both GPC and GNR showing very major errors. Lupetti et al. showed a lower agreement of 94% for erythromycin as well, but observed no very major errors for trimethoprim-sulfamethoxazole. Some other studies on direct methods for AST showed some very major errors for trimethoprim-sulfamethoxazole [15, 16, 18], but only Kerremans et al.  found a very high percentages of very major errors for this antibiotic in GPC, but not in GNR. Therefore, we conclude that the direct Phoenix method using SSTs can be used to reliably report results of AST for GPC, except for trimethoprim-sulfamethoxazole and erythromycin.
The direct method of AST for GNR showed very good agreement with conventional methods for both Enterobacteriaceae and Pseudomonas species, comparable to the routinely used method, with essential agreements and categorical agreements of over 95% for all antibiotics tested (see table 3). Both very major errors occurred with trimethoprim-sulfamethoxazole in Pseudomonas aeruginosa strains that were correctly identified. For these strains, it would never be considered an adequate treatment, due to intrinsic resistance. These errors thus would not have clinical consequences. Funke et al.  also described a categorical agreement of 99.0%, which is comparable with or higher than results from studies on other direct methods of AST [7, 13–16, 26]. Therefore, we conclude that also for GNR, results of the direct Phoenix method for AST can be used to guide antibiotic therapy in bloodstream infections.
The strains tested in this study are a representative sample of the strains most frequently encountered in clinical practice. A limitation of the study is the low number of tested Enterococcus and Pseudomonas strains (3 and 7, respectively), however, both groups show very good agreement, with only few errors.
Inoculating ID and AST broth by using SSTs can be performed as soon as blood culture bottles are taken out of the BACTEC system and takes approximately 30 minutes, whereas a subculture takes up to 24 hours. Therefore, by using the direct method, results of ID and AST can be available up to 23.5 hours earlier than with the routinely used method.