Combination of a Rapid Diagnostic Assay and Antimicrobial Stewardship Intervention for Gram-Negative Bacteremia

Abstract Background Traditional blood cultures for gram-negative bacteremia can take up to 72 hours or more to return results, prolonging the duration of empiric broad-spectrum intravenous antibiotics. The Accelerate Pheno system provides rapid identification and susceptibilities for blood cultures in gram-negative bacteremia. Current data on its clinical utility are mixed overall, so the system requires further research. Methods A multicenter, retrospective quasi-experimental study was conducted comparing the Accelerate Pheno rapid diagnostic system with antimicrobial stewardship intervention and traditional blood cultures alone. Results A total of 264 patients with blood cultures with gram-negative bacteria growth were included in the final analysis (102 pre-intervention, 162 post-intervention). The antimicrobial stewardship team made 364 recommendations in 152/162 (93.8%) patients in the post group. Duration of intravenous therapy was shorter (P < .001) for the post-intervention group (median, 4.0 days) compared with the pre-intervention group (median, 7.5 days). Hospital length of stay was also shorter (P < .001) for the post-intervention group (median, 5.1 days) compared with the pre-intervention group (median, 7.0 days). Readmission rates within 30 days were reduced (P = .042) post-intervention (13.0%) compared with pre-intervention (22.6%). In the post-intervention group, a larger proportion of patients were transitioned to oral therapy at any point (126/162, 77.8%) compared with pre-intervention (62/102, 60.8%; P < .001). Conclusions These results suggest that the Accelerate Pheno system, with active review and intervention by a multidisciplinary antimicrobial stewardship team, is a useful tool in improving both patient-centric and antimicrobial stewardship outcomes.

Effective antimicrobial therapy in patients with bloodstream infections (BSIs) is essential for decreasing morbidity and mortality and helps curb antimicrobial resistance and health care costs [1,2].Gram-negative bacteremia specifically is a major cause of severe illness and death, with 7-and 30-day mortality rates of 8% and 15%, respectively [2,3].An increasing number of these infections are caused by multidrug-resistant organisms, resulting in poor outcomes and difficulty selecting antimicrobial therapy [4,5].Empiric therapy for these infections includes broad-spectrum antibiotics to ensure activity against the most likely causative pathogens and account for drug resistance, with subsequent therapy targeting the specific identified pathogen [6].Appropriate treatment and narrowing of antimicrobial therapy rely on information from blood cultures, including identification (ID) and antimicrobial susceptibility testing (AST) of the organism.Previous studies have shown that traditional blood cultures can take >24 hours to return an ID following initial bacterial growth and up to 48 hours to return AST results [7].As a result of the long turnaround time, clinicians are left treating with broad-spectrum antibiotics for extended durations.Unfortunately, both overuse of broad-spectrum antibiotics and longer durations of therapy are associated with increased rates of antimicrobial resistance and Clostridioides difficile infection (CDI) [8,9].Although antimicrobial stewardship (AMS) intervention after ID and AST results has been shown to be beneficial in multiple ways, including decreasing antibiotic resistance and health care costs, it is still dependent on the time to identification and AST to make appropriate recommendations [1,10].As such, improving time to AST results is of high interest, as this would allow for a quicker and greater impact of AMS, cost savings, prevention of adverse events, and optimization of therapy [1].
Several novel technologies have emerged in recent years to address this delay in accurate ID and AST data [10].One such technology, the Accelerate Pheno system (AXDX; Accelerate Diagnostics, Tucson, AZ, USA), provides ID and AST information for a set panel of common BSI-causing organisms within seven hours of culture positivity [11].This is performed using a combination of automated fluorescence in situ hybridization to identify organisms and morphokinetic cellular analysis to detect phenotypic characteristics [12].Once ID and AST have resulted, the AMS team can utilize the information to ensure rapid and appropriate optimization of antimicrobial therapy.Previous studies have already validated the accuracy and speed of AXDX, with concordance with traditional cultures between 96% and 100% for ID and between 91% and 95% for AST [13][14][15].More recently, AXDX in conjunction with AMS programs has demonstrated reductions in both duration of broad-spectrum antibiotics and total duration of therapy compared with traditional cultures, both of which are metrics that are extremely relevant to combatting growing antibiotic resistance [16][17][18][19][20].
While recent literature has established many advantages of AXDX, further research is needed on its effects on transitioning from intravenous to oral antibiotics, as well as any effects on length of hospital stay [16][17][18][19][20].The current study aims to add to this growing body of literature by analyzing the effects of AXDX in conjunction with an AMS program on the duration of intravenous antibiotics in patients with gram-negative bacteremia.

Study Design
A retrospective, institutional review board-approved, quasiexperimental study was conducted.Data were collected from the electronic medical records of a 719-and a 247-bed hospital, contained within a larger academic health system in Rhode Island, United States, for patients who had blood cultures positive for gram-negative bacteria.In the pre-intervention group, blood was inoculated into Thermo Scientific VersaTREK REDOX1 and REDOX2 bottles (Thermo Scientific, Waltham, MA, USA).Blood culture bottles were then incubated until positive alert, at which point the bottle was gram stained and inoculated to solid media for overnight incubation.Upon overnight incubation, colony identification was performed by Vitek MS (bioMéreieux, Durham, NC, USA), and AST was performed using Vitek 2 AST-GN84 panels (bioMéreieux).ID and AST were then reported via the electronic medical record.There was no routine intervention by AMS on gram-negative bacteremia in the pre-intervention group.In the post-intervention group, blood was inoculated into VersaTREK REDOX1 and REDOX2 bottles, and bottles were incubated until positive alert.Once positive, the bottle was gram stained, inoculated into an AXDX test kit, and inoculated to solid media for overnight incubation.ID and AST from the AXDX test kit were uploaded to the electronic medical record upon completion, and the AMS team was notified via pager.Following overnight incubation, colony identification was performed by Vitek MS, and AST was performed using Vitek 2 AST-GN84 panels; results were uploaded to the electronic medical record.The AMS team reviewed patients for opportunities to optimize, de-escalate, set durations of therapy, and convert antibiotics from intravenous to oral during the hours of 08:00 to 16:30 seven days a week.The AMS team then contacted the primary care team with recommendations for definitive therapy, duration of therapy, and intravenous to oral conversion (if appropriate).For overnight results, the AMS team would review the following morning.

Population and Inclusion/Exclusion Criteria
Eligible patients were those admitted to one of the participating sites who had a positive gram stain for a gram-negative organism contained on the panel of the AXDX.Gram-negative organisms tested by the AXDX Test BC Kit include Klebsiella spp., Citrobacter spp., Enterobacter spp., Proteus spp., Escherichia coli, Serratia marcescens, Pseudomonas aeruginosa, and Acinetobacter baumannii [12].Patients whose blood cultures resulted with gram-negative organisms other than those listed or patients with a prior positive blood culture during the same admission were not included.Patients with polymicrobial BSIs were also not included, as AXDX is unable to reliably report ID and AST for polymicrobial infections [12].Patients were also excluded if they were <18 years of age, if they were discharged before or within 24 hours of AXDX results, were deceased or made comfort measures only before completion of therapy, or if an infectious disease consult was placed before AXDX results (Figure 1).Lastly, any patients who were re-admitted during the study period only had their first admission included in the study as culture data from the previous admission would provide guidance on empiric therapy for the re-admission.
Data were collected via a retrospective chart review of all patients with positive blood cultures with a gram-negative gram stain from December 1, 2017, to September 1, 2018, for the preintervention group, and November 1, 2019, to September 15, 2020, for the post-intervention group.Data points collected included empiric and optimized antimicrobial regimen, time to optimized antibiotic, duration of intravenous and inpatient antibiotics, total duration of antibiotics, and whether any intravenous to oral antibiotic conversions occurred.Safety metrics recorded were readmission within 30 days and development of CDI within 90 days.

Outcomes
The primary outcome of interest was the duration of intravenous antibiotic therapy, measured in days.Duration of intravenous antibiotics was determined starting on the first day of empiric antibiotics until antibiotics were discontinued or until therapy was transitioned to oral antibiotics, whichever came first.If a patient was discharged on intravenous antibiotics, the prescribed duration at discharge was also included in this count.
Secondary outcomes included time to optimal antimicrobial therapy, defined as the discontinuation of broad-spectrum empiric antibiotics and the initiation of targeted therapy based on the susceptibility pattern of the identified pathogen, counted from the time the blood culture was collected to the time an optimized antimicrobial regimen was ordered, in hours; time to discontinuation, in hours, of anti-methicillin-resistant Staphylococcus aureus (anti-MRSA) therapy (vancomycin) and to de-escalation of antipseudomonal beta-lactams (cefepime, piperacillin/tazobactam, and meropenem), counted as the time from start of empiric antibiotics to the time of discontinuation; total duration of antibiotic therapy, in days, including both inpatient and outpatient therapy; rates of conversion from intravenous to oral antibiotics; infection-related length of stay, in days, defined as the time from initial positive culture collection to discharge; total hospital length of stay, in days; incidence of CDI within 90 days of positive culture; and rate of readmission within 30 days of discharge.If antipseudomonal beta-lactams were identified as the optimal treatment agents based on AST results, the cases were excluded from the calculation of time to de-escalation of antipseudomonal beta-lactams.

Statistical Analysis
Continuous variables were represented as medians with interquartile ranges, and categorical variables were represented as number (%).Both baseline characteristics and study outcomes between intervention groups were evaluated utilizing the Wilcoxon rank-sum test for continuous variables and either the Pearson χ 2 test for independence or, if appropriate, the Fisher exact test for categorical variables.
Significant tests for baseline characteristics were further analyzed in post hoc analyses in which critical P values were adjusted for multiple comparisons using Bonferroni's correction.Analyses were performed using Stata software, version 17.0 (StataCorp).Statistical significance was set at a critical P value of ≤.05.

RESULTS
A total of 616 blood cultures positive for gram-negative rodshaped organisms were collected during the study periods.After exclusions, 264 patients were included for analysis, with 102 patients in the pre-intervention group and 162 patients in the post-intervention group.Patient sex assigned at birth, age, and Charlson comorbidity index were similar between the 2 groups (Table 1).The majority of infections were genitourinary in origin (66.3%), followed by gastrointestinal (20.1%).The most common organisms identified were E. coli (61.0%),Klebsiella spp.(21.2%), and Proteus spp.(6.8%).Although the most common causative microbes and sources of infection were similar in both groups, significant differences were observed in the overall distribution of causative microbe and source of infection between groups (Supplementary Tables 1  and 2).
Duration of intravenous therapy was shorter (P < .001) in the post-intervention group (median, 4.0 days) compared with the pre-intervention group (median, 7.5 days) (Table 3).This was paired with a higher rate of conversion from intravenous to oral therapy, with 126/162 (77.8%) converted in the post-intervention group and 62/102 (60.8%) converted in the pre-intervention group (P = .003).Total duration of antibiotic therapy was also shorter (P < .001)for the post-intervention group (median, 13 days) compared with the pre-intervention group (median, 15 days).
Time to optimal antimicrobial therapy was shorter (P < .001)for the post-intervention group (median, 20.2 hours) compared with the pre-intervention group (median, 64.2 hours).In patients receiving anti-MRSA or antipseudomonal beta-lactams, time to appropriate discontinuation of these agents was also reduced in the post-intervention group.Time to discontinuation of anti-MRSA therapy was shorter (P = .002)for the post-intervention group (median, 1.3 days) compared with the pre-intervention group (median, 2.2 days).Additionally, time to discontinuation of antipseudomonal beta-lactams was shorter (P < .001)for the postintervention group (median, 1.5 days) compared with the pre-intervention group (median, 3.4 days).
Length of stay was shorter (P < .001)for the post-intervention group (median, 5.1 days) compared with the pre-intervention group (median, 7.0 days) (Table 3).Similarly, median infectionrelated length of stay was also shorter (P < .001) in the postintervention group (median, 3.9 days) compared with the pre-intervention group (median, 6.0 days).
Regarding safety metrics, there was a numerically decreased but nonsignificant reduction observed between the pre-and post-intervention groups in rates of CDI within 90 days of positive culture (6.9% vs 3.1%, respectively; P = .224)(Table 3).Readmission rates within 30 days were reduced in the postintervention compared with pre-intervention group (13.0%vs 22.6%; P = .042).

DISCUSSION
The present study's goal was to evaluate the clinical impact of a rapid identification and phenotypic AST assay with intervention by an AMS team.Implementation of this process was associated with significant improvements in nearly all measured AMS metrics.More strikingly, we also found reductions in duration of hospitalization and readmission rates in the post-intervention group.A significant reduction in the duration of intravenous antibiotic therapy was observed after the addition of AXDX with intervention by an AMS team.Intravenous therapy is primarily used as initial therapy in the inpatient setting, to quickly acquire adequate systemic levels of antibiotics, or in patients with impaired oral absorption.While these benefits are important in initial management of more severe infections, intravenous therapy also carries risks, including catheter-related infections, thromboembolisms, and extravasation.Early conversion to oral therapy in gram-negative bacteremia is well established to be safe and effective in clinically stable and improving patients with functioning gastrointestinal tracts [21,22].Additionally, other studies have found that conversion to oral medications confers significant cost-savings to the hospital system [23,24].The observed reduction in duration of intravenous therapy among our cohort was likely a result of the combination of active intervention by the AMS team alongside access to quicker identification and AST results via the rapid diagnostic platform implemented here.
Unique to this study, we observed a significant increase in the proportion of patients transitioned to oral antibiotics.Previous studies have either not reported this metric or have not seen a difference in conversion rate [16].The major driver of this outcome is likely the added review and intervention by the AMS team.While Walsh et al. did not see a significant difference in their study, they included active AMS intervention both before and after the implementation of AXDX, and as such AMS was likely already increasing intravenous-to-oral conversion in the pre-intervention group [16].In contrast, our study only included AMS intervention in the post-implementation group, and thus it seems likely that this intervention may have resulted in the observed increased rate of therapy completion with oral antibiotics.Of note, our population in both groups was generally healthy at baseline, with a median Charlson Comorbidity Index of 1, indicating a relatively low occurrence of comorbidities that may complicate conversion to oral therapy.While outpatient parenteral antimicrobial therapy (OPAT) is sometimes necessary, it carries several disadvantages compared with oral antibiotics.Previous literature has shown that OPAT imposes significant cost burdens on patients, as well as reductions in quality of life, compared with oral therapy [25].Up to 9% of patients receiving OPAT will experience a complication with vascular access during their therapy [26].Additionally, OPAT continues to be a challenging option for vulnerable populations, such as people experiencing homelessness and people who inject drugs, due to concerns with the safe administration of parenteral therapy outside of the hospital [27].Increasing rates of completion with oral therapy work to reduce OPAT-associated complications and increase accessibility to effective antimicrobial therapy within these populations.
Compared with the pre-intervention cohort, the duration of both antipseudomonal beta-lactams and anti-MRSA therapy was reduced.The quicker time to de-escalate from antipseudomonal beta-lactams can likely be credited to the faster identification of the causative microbe, as one would be unable to de-escalate this based on the results of gram stain alone.As AXDX is only run on gram-negative gram stains at the included institutions, the reduced duration of anti-MRSA therapy in the post-intervention group was unexpected based on the assumption that clinicians would discontinue anti-MRSA antibiotics as soon as the gram stain resulted as gram-negative, and thus this reduction may have been driven primarily by the AMS intervention as opposed to the rapid diagnostics.We observed a reduction in both infection-related length of stay and total hospital length of stay following the addition of this diagnostic test in conjunction with active intervention by our AMS team.When evaluating length of stay, previous studies have shown mixed results overall [12].One study observed no difference in length of stay in a pediatric population, while another saw a significant reduction in length of stay in patients with gram-negative BSIs [28,29].Bhalodi et al. did not observe any difference when looking at gram-positive and gramnegative BSIs but did note a significant reduction when only looking at gram-negative BSIs [30].Lastly, Dare et al. showed a significant decrease in length of stay for gram-negative BSIs with rapid diagnostics plus AMS with or without real-time notification compared with traditional testing but observed no difference between AMS intervention with and without realtime notification [20].Of the above-described studies, those that showed a reduction in length of stay incorporated active intervention by an AMS team as part of their study, whereas those without significant length of stay reductions did not.Our study further supports that effective utilization of the rapid diagnostics provided by AXDX, when used alongside an AMS team, significantly reduces hospital length of stay for gramnegative BSIs.The differences seen in our study compared with previous studies, both in methodology and results, underpin the importance of the AMS team component when implementing rapid diagnostic testing to achieve the greatest benefit.
There are several limitations present within this study.The pre-and post-intervention design is at risk of confounding due to practice changes over time, through guideline and literature updates as well as institutional changes.Both source of infection and causative microbe were significantly different between the cohorts in the study; however, the most common sources and microbes were the same.Post hoc analyses revealed that the differences in source were driven by the number of genitourinary and central line-associated BSIs between the two groups, while differences in bacterial species were driven by the presence of ten infections by Citrobacter spp., Serratia marcescens, and Acinetobacter baumannii in the preintervention group, with none of these organisms occurring in the post-intervention group (Supplementary Tables 1 and  2).The larger number of nongenitourinary infections may have contributed to worse outcomes in the pre-intervention group, as urosepsis has been shown to have better outcomes than other sources of infection [31].Only readmissions to facilities within the hospital network were able to be captured, and thus a portion of readmissions may have been missed.A large number of patients were excluded; however, the applied exclusion criteria were necessary for the comparison of AXDX with AMS vs traditional cultures alone.Exclusion of polymicrobial/ off-panel organisms, for example, was necessary as AXDX is unable to reliably identify organisms in these scenarios, and thus when a gram stain suggested multiple morphologies the culture was not run on AXDX.Similarly, as we were looking at AMS combined with AXDX, infectious diseases consults were excluded as AMS would no longer be recommending therapy at that point.Lastly, the AMS team reviewed results only between the hours of 08:00 and 16:30, which may have limited the benefit of the rapid turnaround time of AXDX.Real-time review of results 24 hours per day may provide improved results; however, this is unlikely to be feasible due to staffing limitations at most institutions.
While this study demonstrated improvement in several measurable clinical outcomes associated with the use of AXDX, several metrics remain to be evaluated.Although it can be assumed that reducing length of hospital stay is likely to provide significant economic benefits, the degree of this benefit would need to be weighed against the costs of AXDX and further evaluated in follow-up studies to fully quantify.A significant reduction in readmission rate in the post-intervention group compared with the pre-intervention group was observed as well; however, our study sample size may not have been large enough to capture the true event rate between groups.

CONCLUSIONS
Rapid diagnostic tests in conjunction with intervention by an AMS team for gram-negative BSIs offer many advantages over traditional cultures alone.This study supports previously observed conclusions of decreased time to optimal therapy, reduced length of stay, and reduced duration of antibiotic therapy.In addition to these findings, this study also showed a reduction in duration of intravenous therapy, alongside an increased rate of conversion to oral antibiotics compared with traditional cultures alone.These differences likely reduce costs, increase patient quality of life, and improve access to care.Overall, these results suggest a definite clinical utility associated with integration of AXDX in conjunction with review by an AMS team as compared with the use of a traditional automated blood culture system without AMS intervention.

Figure 1 .
Figure 1.Flowchart representation of included and excluded cases.a Cultures with organisms unable to be identified by the Accelerate Pheno system were excluded from the pre-intervention group.Abbreviations: AST, antimicrobial susceptibility testing; CMO, comfort measures only; ID, infectious disease.

Table 1 . Baseline Characteristics
Age and Charlson Comorbidity Score are reported as median (IQR).Sex, presumed source, and isolated bacteria are reported as number (%).P values were calculated using the Wilcoxon rank-sum test for continuous variables and either the Pearson χ 2 test or the Fisher exact test for categorical variables.Statistically significant values are in bold (P ≤ .05).Abbreviations: CLABSI, central line-associated bloodstream infection; IQR, interquartile range; SSTI, skin and soft tissue infection.a Pneumonia, neutropenic fever, graft infection, spontaneous bacterial peritonitis, unknown source.b Citrobacter species, Acinetobacter baumannii, Serratia marcescens.

Table 2 . Recommendations Made by the Antimicrobial Stewardship Team in the Post-intervention Group
Abbreviations: IV, intravenous; PO, per os.

Table 3 . Measured Outcomes
Data are reported as either No. (%) or median (IQR).P values were calculated using the Wilcoxon rank-sum test for continuous variables and either the Pearson χ 2 test or the Fisher exact test for categorical variables.Statistically significant values are in bold (P ≤ .05).Abbreviations: CDI, Clostridioides difficile infection; IQR, interquartile range; IV, intravenous; LOS, length of stay; MRSA, methicillin-resistant Staphylococcus aureus.