Among Escherichia coli and Klebsiella pneumoniae isolates in the United States, 85.4% of extended-spectrum beta-lactamases (ESBLs) were due to bla _CTX-M. ^{Reference Castanheira, Kimbrough, DeVries, Mendes and Sader1} With the increasing incidence of ESBL-producing isolates, rapid diagnostic tests, which include genotypic information such as the detection of bla _CTX-M, have been shown to improve time to more appropriate, and even optimal, empiric therapy. ^{Reference Tamma, Komarow and Ge2} This is critical since patients with ceftriaxone-resistant infections have been found to have worse outcomes. ^{Reference Tamma, Komarow and Ge2,Reference Duffy, Karlsson and Reses3} The BioFire BCID2 is a second-generation multiplex panel which rapidly identifies 30 different Gram-negative, Gram-positive, and yeast pathogens along with 10 antimicrobial resistance genes from blood cultures. ⁴ Specifically, the bla _CTX-M resistance marker detects the presence or absence of the most common family of ESBL enzymes; however, little is known regarding the safety of de-escalation in the absence of this marker. ^{Reference Pogue, Heil and Lephart5} Genotypic markers of ceftriaxone resistance appear to outperform ESBL clinical prediction rules. ^{Reference Andrews, Timbrook, Fisher and Tritle6} Thus, we created a genotypic antibiogram to use in combination with BCID2 results to assist with improving our institutional guidance for empiric antimicrobial therapy selection for patients with Enterobacterales bloodstream infections (BSI).

For this study, all positive BCID2 results with monomicrobial Enterobacterales BSIs at our academic medical center from 8/1/2021 to 11/1/2022 were retrospectively reviewed. Isolates with multiple positive resistance markers were excluded. Patient characteristics, BCID2, culture, and susceptibility results were collected. Immunocompromised patients were defined as people living with HIV with CD4 < 200, having received a solid organ or bone marrow transplant, or undergoing treatment for a hematologic and/or oncologic malignancy. Community-onset infections were defined as blood cultures collected < 48 hours from hospital admission. We performed descriptive statistics for cohort characteristics. Sensitivity, specificity, positive predictive values (PPV), and negative predictive value (NPV) of the bla _CTX-M marker was compared to ceftriaxone susceptibility. Antimicrobial susceptibility testing was performed using the MicroScan WalkAway System (Beckman Coulter) and the MicroScan Negative MIC 56 antimicrobial panels, according to the manufacturer’s instructions. ESBL and AmpC producers were identified per our institutional protocol (Supplemental Material). CLSI guidance was followed for interpreting results (M100) and creation of the antibiogram (M39) utilizing the first patient isolate per year methodology. ⁷ Our institutional review board deemed this a quality improvement project exempt from review.

Over 15 months, 455 Enterobacterales bloodstream isolates were identified from 452 unique patients. Of those, 236 (52%) were male patients, 189 (41%) immunocompromised, and 342 (75%) had community-onset infections. The most common species identified were Escherichia coli (55%) and Klebsiella pneumoniae group (17%). bla _CTX-M was detected in 48 (11%) isolates from 41 (85%) E. coli and 6 (13%) K. pneumoniae group (Supplemental Table 1). E. coli and K. pneumoniae isolates that did not harbor bla _CTX-M detected were 97% and 100% susceptible to ceftriaxone, respectively. bla _KPC was detected in 1 isolate (K. variicola), and excluded from our analysis. Additionally, no other carbapenemase genes were detected. For E. coli, bla _CTX-M sensitivity and specificity for detection of ceftriaxone resistance was 85% and 100%, respectively; bla _CTX-M PPV for ceftriaxone resistance was 100%, while the NPV of absent bla _CTX-M for ceftriaxone susceptibility was 97% (Table 1). Of 7 bla _CTX-M negative E. coli isolates, 6 were identified as ESBL and one as an AmpC producer based on phenotypic lab protocols. For K. pneumoniae group, bla _CTX-M sensitivity and specificity for detection of ceftriaxone resistance was 60% and 100%, respectively; bla _CTX-M PPV for ceftriaxone resistance was 100%, while the NPV of absent bla _CTX-M for ceftriaxone susceptibility was 94% (Table 1). Among the other Enterobacterales species detected on BCID2, only one had bla _CTX-M detected despite 24% (31/129) having ceftriaxone resistance. Figure 1 describes susceptibilities to common antimicrobials, delineated on the presence or absence of BCID2 resistance markers, for E. coli and K. pneumoniae group since these organisms are most likely to harbor bla _CTX-M, whereas other organisms are more likely to have other mechanisms of ceftriaxone resistance (eg, Enterobacter species and AmpC). ^{Reference Castanheira, Kimbrough, DeVries, Mendes and Sader1} Supplemental Figure 1 describes susceptibilities for all organisms.

Table 1. BCID2 organisms with bla _CTX-M detected sensitivity, specificity, positive, and negative predictive values

Note. SN, sensitivity; SP, specificity; PPV, positive predictive value; NPV, negative predictive value.

^a All organisms with bla _CTX-M detected on BCID2. Those without bla _CTX-M detected are not shown.

^b Includes Klebsiella pneumoniae, Klebsiella quasipneumoniae, and Klebsiella variicola.

^c Organism cultured that was bla _CTX-M positive identified as Providencia rettgeri.

Figure 1. Genotypic Antibiogram.

To our knowledge, this study is the first to assess the utility of BCID2 to create a genotypic antibiogram. Previous literature focused on the Verigene Gram-negative blood culture nucleic acid test’s ability to provide empiric treatment recommendations based on the presence or absence of resistance markers. ^{Reference Pogue, Heil and Lephart5,Reference Spafford, MacVane and Humphries8} Our approach is novel in assessing BCID2, which contains additional resistance markers and organism targets in comparison to Verigene, providing clinicians with additional information to target antimicrobial therapy. ⁴ Results from this study demonstrate similar PPV and NPVs for ceftriaxone susceptibility. ^{Reference Pogue, Heil and Lephart5,Reference Rödel, Karrasch and Edel9} Based on the high NPV in E. coli and K. pneumoniae group, clinicians can confidently de-escalate therapy to ceftriaxone when the bla _CTX-M resistance marker is not detected at our institution. A genotypic antibiogram (Figure 1) was distributed to clinicians simultaneously with our updated BCID2 utilization guidance document ¹⁰ which provides empiric antibiotic recommendations based on the BCID2 results. While rapid diagnostics are useful in identifying organisms to target, clinicians often struggle to interpret these complex tests, and education and guidance on how to optimally utilize and interpret the results is continually needed. ^{Reference Donner, Campbell, Lyden and Van Schooneveld11} A recent study ^{Reference Hasegawa, Livorsi, Perencevich, Church and Goto12} questioned the overall efficacy and diagnostic accuracy of cumulative antibiograms’ ability to predict resistance for isolates; however, genotypic blood culture antibiograms likely have improved predictive capability.

Limitations include the retrospective, single-institution design; however, this project is easily replicable and can provide other facilities with a framework for assessing their specific resistance patterns and rapid diagnostic data. Extrapolation of the results to other multiplex panels (eg, pneumonia panel) should be done cautiously. Additionally, there may be limited value of an annual genotypic antibiogram given the limited sample size and additional cost/time associated with its creation. However, genotypic antibiograms may provide insight into the monitoring of epidemiologic trends in ESBLs within an institution. Other limitations include the low number of bla _CTX-M isolates, which limits applicability to only a select number of organisms. Since the proportion of bla _CTX-M isolates was representative of the overall BCID2 results, the genotypic antibiogram still provides clinically relevant information regarding susceptibilities. Further, for our institution, absence of bla _CTX-M does not reliably predict ceftriaxone susceptibility for non-E. coli or non-K. pneumoniae isolates. Finally, antimicrobial utilization and prescriber response based on BCID2 results was not collected; however, previous literature supports that utilization of rapid diagnostics, when coupled with stewardship, reduces time to optimal antimicrobial therapy, and our stewardship program reviews BCID results daily. ^{Reference Banerjee and Humphries13} Further studies should be performed to explore the utilization and impact of BCID2 gram-negative resistance markers on time to optimal therapy and patient outcomes.

This study demonstrated that the majority of ceftriaxone-resistant E. coli and Klebsiella BSI harbor bla _CTX-M, and carbapenemases are rare in our institution. Creating a genotypic antibiogram assisted in providing updated guidance for our clinicians on treatment of Enterobacterales bloodstream infection.