Clinical Performance of the Novel GenMark Dx ePlex Blood Culture ID Gram-Positive Panel

Study population.

The study population included patients of all ages and genders with positive blood cultures collected for standard patient care and diagnosis at 10 geographically diverse clinical sites located in the United States from the following nine cities: Albuquerque, NM; Baltimore, MD; Charleston, SC; Danville, PA; Detroit, MI (two sites); Harvey, IL; Indianapolis, IN; Milwaukee, WI; and San Diego, CA.

Specimen collection.

A total of 719 samples with positive Gram stain results were prospectively collected; 400 samples were collected from June 2014 through July 2016 from three sites and frozen for future testing, and 319 samples were collected from January through March 2018 from five sites and tested fresh. The majority of samples were collected from subjects at hospitals or medical centers. To supplement the results of the prospective collection, 586 samples with positive Gram stain results for lower-prevalence Gram-positive organisms were retrospectively collected from 10 sites. Additionally, 566 samples were contrived for Gram-positive organisms with very low prevalence. For evaluation of the Pan targets only, 484 retrospective samples with Gram-negative or fungal Gram stain results were included.

Contrived samples were prepared by aseptically injecting 3 to 10 ml of human whole blood (Bioreclamation, Westbury, NY) into a BD Bactec blood culture bottle (Plus Aerobic/F, Plus Anaerobic/F, Lytic/10 Anaerobic/F, or Peds Plus/F). Table S1 in the supplemental material contains a list of organisms used for the contrived samples. The 27 Micrococcus samples were contrived using quantified glycerol stocks spiked into incubated blood. For the remaining samples, the bottles were then inoculated with a colony or colonies from a pure culture. The bacterial suspension was prepared by using a loop to collect fresh culture from an agar plate and suspending it in saline to approximately a 0.5 McFarland standard via reading the optical desity at 600 nm (OD₆₀₀) (0.5 McFarland is approximately 1.5 × 10⁸ CFU/ml). This bacterial suspension was diluted either 1:100 (n = 483), 1:1,000 (n = 27), or 1:10,000 (n = 29), and then either 100 μl (n = 499) or 1 ml (n = 40) was used to inoculate a Bactec bottle containing blood. The 1-ml inoculum was used for some of the Bacillus and Cutibacterium acnes replicates. The inoculum volumes were determined based on successful growth and time to detection in preliminary studies. The CFU added to the bottles ranged from ∼1 × 10³ to 2 × 10⁶. The time to detection varied from 7 h to 5 days in 97% of the samples and was up to 10 days for the remaining samples. Of note, the organisms that required more than 5 days were Bacillus spp., Lactobacillus spp., Corynebacterium spp., and C. acnes.

Overall study design and conduct.

Prospectively and retrospectively collected samples were cultured and tested as ordered, per the sites’ standard operating procedures. All samples were deidentified, residual positive blood culture samples provided after the standard-of-care (SOC) testing was completed. The study was performed using two protocols, one for collection only and a second for collection and testing; both were approved by a central institutional review board (IRB) (Quorum, Seattle, WA) and/or the site’s IRB.

Samples were tested at one of five clinical sites with the GenMark Dx ePlex IUO Blood Culture Identification Gram-Positive (BCID-GP) Panel. Investigational results were compared to the sites’ SOC procedures for organism identification and sequencing for antimicrobial resistance genes (ARGs).

GenMark Dx ePlex BCID-GP Panel testing.

The BCID-GP Panel was tested within 12 h of positivity as instructed by the manufacturer using the materials in the kit. Briefly, after the blood culture bottle was inverted several times to mix, 50 μl was removed and added to the BCID-GP Panel cartridge through the sample loading port and the cap was depressed to close the port. Each cartridge was barcoded according to the manufacturer’s instructions, scanned at the ePlex instrument, and inserted into an available bay. Upon completion of the assay run, the ePlex instrument ejected the cartridge for disposal and a BCID-GP Panel report was generated.

The BCID-GP Panel runs on the ePlex instrument, which automates all aspects of nucleic acid testing, including extraction, amplification, and detection, combining electrowetting and GenMark’s eSensor technology in a single-use cartridge. eSensor technology is based on the principles of competitive DNA hybridization and electrochemical detection as previously described (20).

The BCID-GP Panel identifies 20 Gram-positive bacteria and four antimicrobial resistance genes (Table 1). It should be noted that if there are multiple Gram-positive cocci in a sample that is also positive for a resistance marker, there is no way to link the gene to the detected species. This would likely most commonly occur with mecA when the sample contains Staphylococcus aureus and one or more other staphylococci. The BCID-GP Panel also contains two targets to detect non-Gram-positive organisms using a Pan target approach. The Pan Gram-Negative target detects up to 95% of Gram-negative bacteria, including (but not limited to) Enterobacterales, Acinetobacter, Pseudomonas, Bacteroides, Stenotrophomonas maltophilia, Serratia, and Neisseria. Four key Candida species (Candida albicans, C. glabrata, C. krusei, and C. parapsilosis) are detected by the Pan Candida target. Like other commercial syndromic panel tests for use on positive blood culture bottles, this panel is designed to be a useful, rapid adjunct for the diagnosis of bloodstream infections.

Comparator methods.

The comparator methods for organism identification consisted of standard laboratory procedures, including traditional culture, FDA-cleared MALDI-TOF MS (i.e., bioMérieux Vitek MS or Bruker Biotyper), and automated microbiological and biochemical techniques (e.g., Becton, Dickinson [BD] Phoenix, bioMérieux Vitek 2, or Siemens MicroScan). The majority (73.4%) of prospective fresh samples had organisms identified by an FDA-cleared MALDI-TOF MS system (Vitek MS or Bruker). The versions used by the sites are as follows: Johns Hopkins, the Bruker Microflex LT/SH IVD version 2.0; Detroit Medical Center, the Bruker Microflex LT/SH IVD version 2.0; Medical University of South Carolina (MUSC), Bruker Microflex LT/SH IVD version 4.0.11.0; Indiana University, Bruker Microflex LT/SH IVD version 3.2.14; Henry Ford Health System, Vitek MS version 2.2; and Tricore Laboratories, Bruker Microflex LT/SH IVD version 3.2.12.2. Prospective frozen samples were comprised of organisms identified by the bioMérieux Vitek 2 microbiology system (25.8%), Siemens MicroScan (23.3%), BD Phoenix (2.5%), or a wide variety of conventional manual methods. Organisms in the retrospective Gram-positive samples were identified primarily by the BD Phoenix (44.2%) or bioMérieux Vitek 2 (17.4%) microbiology system. Methods differed between the fresh and frozen samples due to sites changing methods between the collection time periods.

Due to known issues with organism identification, samples with Corynebacterium, Staphylococcus epidermidis, Staphylococcus hominis, or C. parapsilosis identified by standard laboratory procedures were confirmed using analytically validated PCR amplification assays followed by bidirectional sequencing (PCR/sequencing) or 16S rRNA gene sequencing (21–24).

The comparator methods for ARGs were analytically validated real-time PCR (qPCR) amplification assay(s) followed by bidirectional sequencing for samples with the associated organisms identified (i.e., Staphylococcus and Enterococcus).

Method for resolution of discordant results.

Results from collected samples that were discordant between the BCID-GP Panel and the comparator method(s) (i.e., false negative or false positive) were tested with PCR/sequencing and/or 16S sequencing to determine the presence or absence of the organism. For organisms and ARGs included on the BCID-GP Panel, a false-negative result occurred when the comparator method identified an organism or ARG that was not detected with the BCID-GP Panel. A false-positive result occurred when the BCID-GP Panel detected an organism or ARG that the comparator method did not identify.

Statistical methods.

Positive percent agreement (PPA) and negative percent agreement (NPA) with comparator method results were determined for each target detected by the BCID-GP Panel. PPA was calculated as 100 × number of TP/(number of TP + number of FN), and NPA was calculated as 100 × number of TN/(number of TN + number of FP), where TP is true positives, FN is false negatives, TN is true negatives, and FP is false positives. The two-sided 95% score confidence interval (CI) was calculated for PPA and NPA.

Demographic/sample information, sample disposition, and run/sample accountability.

Demographic information for the evaluable prospectively and retrospectively collected subject samples is provided in Table S2. Slightly more than 50% of the subjects were male, with a median age of 60 years.

Thirteen different blood culture bottle types from three manufacturers were used. Of the 10 clinical sites collecting patient samples, 6 used BD Bactec, 2 used Thermo Fisher Scientific VersaTREK, and 2 sites used both BD and bioMérieux BacT/Alert. There were no discernible trends for a specific bottle type (data not shown). Ninety percent of samples were either tested or frozen within 12 h of blood culture bottle positivity. The prevalence of BCID-GP Panel targets by age group during prospective collection is provided in Table S3.

A total of 719 prospectively collected, 1,070 retrospectively collected (includes both the Gram-positive samples and the non-Gram-positive samples) and 566 contrived samples were tested with the BCID-GP Panel, including 10 withdrawn samples (7 prospective and 3 retrospective) tested prior to withdrawal. Of the 2,354 samples initially tested with the BCID-GP Panel, 108 yielded invalid results, for an initial validity rate of 95.4%. After repeat testing, only 2 samples (1 contrived and 1 retrospective) of the 108 repeated as invalid, for a final validity rate of 99.9%. There were 1,297 evaluable clinical samples (including prospective and retrospective samples) and 565 contrived samples for the Gram-positive targets, as well as 1,777 evaluable clinical samples for the Pan targets.

Panel performance.

There were 1,297 positive blood culture samples with Gram stains that displayed Gram-positive organisms, including 1,220 with only Gram-positive, 64 with mixed Gram-positive and Gram-negative organisms, 3 with mixed Gram-positive organisms and yeast, 3 with mixed Gram-positive organisms, Gram-negative organisms, and yeast, and 7 with Gram-variable organisms. The BCID-GP Panel correctly identified the organisms/ARGs identified by comparator methods in 93% of these samples. The overall sample accuracy where the BCID-GP Panel and comparator methods detected the same organisms/ARGs was 89% before resolution of discordant results. After correction for organisms/ARGs missed by comparator methods but detected by the BCID-GP Panel and confirmed by other methods, the overall sample accuracy was 91%. The overall weighted PPA across all 26 targets was 96%. These results exclude the 28 Gram-positive (2.2%) organisms representing 12 genera identified by comparator methods but not targeted by the BCID-GP Panel (Table S4). These 28 isolates included 6 Clostridium spp., 6 Rothia spp., 3 Aerococcus isolates, 2 Microbacterium isolates, 2 Peptostreptococcus spp., and single isolates representing 7 different taxa and 2 organisms not identified beyond Gram stain morphology.

Positive percent agreement and NPA with 95% CI of the BCID-GP Panel targets with comparator methods are provided in Table 2 for Gram-positive targets. Table S5 details the performance of the BCID-GP Panel for all samples stratified by the species detected by the comparator methods or in the known contrived samples. Table S6 summarizes the BCID-GP Panel clinical performance after resolution of discordant results. Specific details of the discordant results are provided throughout the narrative summaries below.

In the clinical arms of the study, only 12 Bacillus cereus group organisms were recovered by SOC methods (Table 2). For those 12 samples, 11 were correctly identified by the BCID-GP Panel (91.7% PPA) and the NPA was 100%. Overall results with the contrived samples included a PPA of 98.3% and NPA of 100%.

A total of 334 enterococci, including 191 Enterococcus faecalis samples and 125 Enterococcus faecium samples, were tested with an overall PPA of 97.6% and NPA of 99.9%. E. faecalis was more frequently isolated from clinical samples (n = 139) than E. faecium (n = 65). Enterococcus was detected in one false-positive prospective sample using PCR/sequencing. When considering the major species separately, the overall PPA for E. faecalis was 95.8% and the NPA was 100% (Table 2). E. faecalis was not detected in four of the eight false-negative clinical samples using PCR/sequencing. Three contained E. faecium, while Lactococcus lactis was isolated from one sample. For E. faecium, the overall PPA was 98.4% and the NPA was 99.5% (Table 2). Five of the eight false-positive samples were detected by PCR/sequencing as E. faecium.

Only two Listeria clinical samples were identified by SOC methods, so 75 contrived samples containing several Listeria species (i.e., Listeria innocua, L. ivanovii, L. monocytogenes, L. seeligeri, and L. welshimeri) were included to supplement the data. The overall PPA and NPA for the Listeria target were 98.7% and 99.9%, respectively. The two clinical samples grew L. monocytogenes; therefore, 46 of 48 of the samples for this organism were contrived. For the L. monocytogenes target, the two clinical samples were detected by the BCID-GP Panel and the assay accurately detected all contrived samples as noted in Table 2.

As expected, Staphylococcus targets were abundant during the clinical arms of the study, with a total of 647 staphylococci detected by SOC testing. Eighteen species or subspecies of staphylococci were included (Table S5). The BCID-GP Panel detected 632 of the staphylococci, with a PPA for the clinical samples of 97.7%; the NPA was 98.5% (Table 2). Staphylococcus was not detected in five false-negative samples using PCR/sequencing. These samples contained a variety of Gram-negative and Gram-positive organisms (one contained Escherichia coli, one Klebsiella pneumoniae, and one S. salivarius, and two had no other organisms detected). For the 10 false-positive samples, Staphylococcus species were detected in 9 by PCR/sequencing. All 105 contrived samples (59 S. aureus, 1 S. epidermidis, and 45 S. lugdunensis) were accurately identified by the BCID-GP Panel.

There were 291 total S. aureus isolates recovered by SOC methods, with 98.8% overall accuracy by the BCID-GP Panel. The PPA was 96.9% for all clinical samples (prospective and retrospective), and the NPA was 99.4% (Table 2). There were nine false-negative samples compared to SOC methods. S. aureus was not detected by PCR/sequencing in three false-negative prospective samples, but Staphylococcus simulans, Streptococcus agalactiae, and Klebsiella pneumoniae were detected. However, for five of the six false-positive samples, S. aureus was detected using PCR/sequencing. All 59 contrived samples were accurately identified by the BCID-GP Panel, for a PPA and NPA of 100% (Table 2).

The data for S. epidermidis were less robust than the S. aureus data but still in the acceptable range (PPA > 90%). A total of 159 S. epidermidis isolates were detected by SOC methods. The overall PPA and NPA (including the singular contrived sample) were 93.1% and 98.8%, respectively. There were six false-negative samples that contained a variety of different staphylococcal species as per testing by PCR/sequencing, including four samples identified as S. aureus (originally misidentified as S. epidermidis by the SOC method). The other staphylococci identified were Staphylococcus capitis and Staphylococcus pettenkoferi.

Standard-of-care methods detected six S. lugdunensis-positive blood cultures, and all six were correctly identified by the BCID-GP Panel. All 45 contrived samples were also accurately detected. There were two false positives among the clinical samples which, when analyzed by PCR/sequencing, were found to have S. lugdunensis present. Overall, the PPA was 100% and the NPA was 99.8% (Table 2).

The BCID-GP Panel detects the Streptococcus genus, with the ability to distinguish four species or groups—S. agalactiae, S. anginosus group, S. pneumoniae, and S. pyogenes. Sixteen species or groups within the Streptococcus genus were detected during the study (Table S5). The PPA ranged from 95.6% for the S. anginosus group to 98.1% for S. pyogenes. Similar to the data for the other Gram-positive targets, the NPAs were high, ranging from 99.4% to 100% (Table 2). There were 13 samples with 16 false-positive results across the five Streptococcus targets, whereby the respective organism was detected by PCR/sequencing in 11 (8 Streptococcus species, 1 S. agalactiae, 1 Streptococcus intermedius, and 1 S. pneumoniae). Of the 16 samples that yielded 18 false-negative results, 4 samples had no streptococci detected. Streptococcus mitis was detected by PCR/sequencing in one false-negative S. agalactiae sample, one isolate each of Granulicatella adiacens, S. dysgalactiae, and S. lutetiensis was detected in three false-negative S. anginosus group samples, and of three false negatives for S. pneumoniae, two were positive for S. mitis and one was positive for S. anginosus by PCR/sequencing.

The data for the Pan targets are presented in Table 3. No contrived samples with Gram-negative or fungal organisms were tested. One of the four Candida species contained in the Pan Candida portion of the assay was detected by SOC methods in a total of 105 clinical samples. The overall PPA was 92.4%, and the NPA was 99.9% (Table 4). Recall that clinical samples without positive Gram stain results were included to increase the sample size to evaluate the Pan targets. Of the eight false-negative samples, only two had positive Gram stain results intended to be tested with the BCID-GP Panel. Notably, the Gram stain only indicated Gram-positive organisms and missed the presence of yeasts that were later identified, with one sample having C. glabrata verified by PCR/sequencing. The remaining six false-negative samples had fungal and/or negative Gram stain results, of which four had C. albicans verified by PCR/sequencing. C. glabrata was detected in one of two false-positive samples using PCR/sequencing. A Gram-negative organism was identified in 441 clinical samples, and the Pan Gram-Negative target demonstrated an overall PPA of 95.7% and NPA of 99.6%. Of the 19 false-negative samples, only 8 had positive Gram stain results intended to be tested with the BCID-GP Panel, and two had Gram-negative organisms verified by PCR/sequencing. The remaining 11 false-negative samples had negative Gram stain results, of which 9 had Gram-negative organisms verified by PCR/sequencing. By the comparator methods, the organisms that were not detected included the following: one Achromobacter xylosoxidans subsp. xylosoxidans isolate, two Acinetobacter baumannii isolates, two Bacteroides fragilis isolates, four Bacteroides thetaiotaomicron isolates, one Campylobacter gracilis isolate, one Delftia acidovorans isolate, two K. pneumoniae isolates, one Ochrobactrum anthropi isolate, one Proteus mirabilis isolate, one Providencia stuartii isolate, one Pseudomonas aeruginosa isolate, and one isolate each of Wolinella and Veillonella. Of note, the assay correctly identified three samples containing Salmonella enterica subsp. enterica serovar Typhi. Of seven false-positive samples, two samples had E. coli and K. pneumoniae, respectively, detected, whereas in the other five samples no Gram-negative organisms were detected using PCR/sequencing (Table 3). Of the 1,220 samples with positive Gram stain results, 11 (0.9%) were coinfected with a Gram-negative bacterium or Candida detected by either the BCID-GP Panel (n = 6), SOC (n = 3), or both (n = 2). The BCID-GP Panel alone detected five Gram-negative organisms and one Candida organism, with PCR/sequencing confirming one to be K. pneumoniae. Both methods detected Gram-negative E. coli in two samples. In 3 samples, SOC alone detected Veillonella, C. glabrata, and C. parapsilosis.

Table 4 contains the data for the antimicrobial resistance gene targets. Similar to the case with other commercially available highly multiplexed blood culture identification tests, if there are multiple Gram-positive cocci in a sample that is mecA positive, there is no way to link the gene to the detected species if they are all staphylococci (19, 25, 26). There were 413 mecA-positive staphylococci detected in the clinical study by the comparator methods, and 11 contrived specimens were also tested. The overall PPA and NPA were 97.2% and 96.6%, respectively. Among the clinical specimens, there were 12 false-negative samples. However, additional testing indicated that two of the false-negative samples were contaminated with mecA during the original extraction process for the comparator method testing. Samples were reextracted twice and upon repeat analysis, mecA was not present. Eleven false-positive mecA samples were also noted, and in four of these (three prospective and one retrospective), mecA was found to be present using an FDA-cleared multiplex PCR assay. Among the clinical samples, 194 Staphylococcus aureus isolates were mecA positive by comparator methods, with 190 mecA genes detected by the BCID-GP Panel (PPA, 97.9%). Up to two coagulase-negative Staphylococcus species (CoNS) were detected in 154 samples that were mecA positive by comparator methods (BCID-GP Panel detected mecA in 149; PPA, 96.8%). These included S. epidermidis (n = 98), S. capitis (n = 7), S. haemolyticus (n = 5), S. hominis (n = 27), S. auricularis (n = 1), S. saprophyticus (n = 1), S. simulans (n = 1), S. epidermidis/S. hominis (n = 9), and one isolate each of S. epidermidis/S. capitis, S. epidermidis/S. lugdunensis, S. capitis/S. hominis, S. cohnii/S. hominis, and S. haemolyticus/S. hominis (data not shown).

This is the first commercial panel to detect mecC among the staphylococci. Interestingly, there were no mecC-containing staphylococci detected during the clinical study, confirming reports of low prevalence of mecC in the United States (27, 28). The assay detected mecC in all of the 49 contrived samples tested (Table 4).

The major vancomycin resistance determinant among the clinical enterococci was vanA (n = 65). In terms of species distribution for vanA-positive samples, E. faecium accounted for 75% and E. faecalis for the majority of the others. Among the 65 enterococcal samples for which comparator methods detected vanA, 13 had E. faecalis, 49 had E. faecium, 2 samples had dual E. faecalis/E. faecium, and 1 contained Enterococcus avium. The BCID-GP Panel detected vanA in 61/65 samples (PPA, 93.8%). There were three E. faecalis isolates and one E. avium isolate that were determined to be vanA positive by comparator methods but were missed by the BCID-GP Panel. The clinical data were supplemented with 60 contrived samples (Table S1). Overall, the BCID-GP Panel identified 121 of the 125 (96.8%) vanA-positive samples. The overall NPA was 99.0%. For two of the false-negative samples, a vanA signal was present, but without the presence of an associated Enterococcus organism detected by the BCID-GP Panel, vanA was not reported. Similarly, additional testing indicated that two of the false-negative samples were contaminated with vanA during the original extraction process for the comparator method testing. Samples were reextracted twice and vanA was not present. When tested using the FDA-cleared multiplex PCR assay mentioned above, one of these samples did not have vanA detected. vanA was also detected in one of the two false-positive samples that were tested with the same FDA-cleared multiplex PCR assay (Table 4). There was only one vanB-positive sample containing E. faecalis among the clinical samples tested; the assay accurately detected vanB in that sample and all 52 contrived samples.

Contamination rule-out targets.

The BCID-GP Panel includes five targets, Bacillus subtilis group, Corynebacterium, C. acnes, Lactobacillus, and Micrococcus, where assay design favored very high specificity over sensitivity because the primary clinical utility of these targets is to rule out contamination (Table 5). For these five targets, the NPA in clinical samples and overall in all samples was at least 99.8%. For the Corynebacterium target, comparator methods identified 14 different species (Table S5). Of 11 samples with false-negative results, 7 were not identified to the species level by the laboratories performing the assay, resulting in an overall PPA of 84.5%. The other four species that were not detected by the BCID-GP Panel were C. afermentans, C. jeikeium, C. pseudotuberculosis, and C. urealyticum.

An investigation into these 11 false-negative results was performed using PCR/sequencing and 16S rRNA sequencing. For five samples, PCR/sequencing was negative and 16S rRNA sequencing detected C. acnes, Lactobacillus fermentum, Macrococcus caseolyticus, and S. pettenkoferi (the fifth sample was not tested with 16S rRNA sequencing). Five samples had Corynebacterium detected by PCR/sequencing and/or 16S rRNA sequencing. The remaining sample had indeterminate results. Corynebacterium was detected in two false-positive clinical samples using PCR/sequencing.

For C. acnes, the overall PPA was 93.5%. PCR/sequencing detected C. acnes in one false-positive clinical sample. In the two false-negative samples, PCR/sequencing detected C. acnes.

Lactobacillus was detected in all of the positive clinical samples, and there was only one false-negative contrived sample, for an overall PPA of 97.8% (45/46). There were two false-positive samples, and PCR/sequencing detected Lactobacillus casei in one sample.

Forty-four clinical samples contained Micrococcus species; the assay failed to detect five of them, four of which were identified to the genus level only by the comparator methods, and the fifth one was identified as Micrococcus luteus/Micrococcus lylae (data not shown). Micrococcus was not detected in three false-negative clinical samples using PCR/sequencing; one Brevibacterium ravenspurgense isolate, one Nesterenkonia halotolerans isolate, and one S. pettenkoferi isolate were detected. Consistent with the performance with the other contamination targets, the NPA overall was close to 100%.

Table S7 summarizes single infections with discordant results, that is, BCID-GP Panel false-negative or false-positive results compared to the comparator methods. The results for mixed infections are described in detail below, and concordant mixed infections are highlighted in Table 6; discordant mixed infections are catalogued in Table S8.

Mixed infections.

Overall, 184 mixed infections were identified by comparator methods and/or the BCID-GP Panel. Codetections of on-panel targets identified by comparator methods that were concordant with the BCID-GP Panel in clinical samples are provided in Table 6. The BCID-GP Panel correctly identified on-panel targets in 85 (46%) samples. Of these, 70 (82%) had two organisms, 14 (16%) had three organisms, and 1 (%) had four organisms. Of the 184 codetected samples, 99 (54%) discordant samples are provided in Table S8 and consist of codetections that included an organism/ARG not detected by the BCID-GP Panel (i.e., false negative) in 58 samples or an organism/ARG that was detected by the BCID-GP Panel but not identified by comparator methods (i.e., false positive) in 65 samples (some samples had both false-negative and false-positive detections). Thirty-one (48%) of the 65 false-positive detections were confirmed and 23 (40%) of the 58 false-negative detections were not confirmed by methods for resolving discordant results.