Evaluating the impact of rapid antimicrobial susceptibility testing for bloodstream infections: a review of actionability, antibiotic use and patient outcome metrics

Abstract

Antimicrobial susceptibility testing (AST) is a core function of the clinical microbiology laboratory and is critical to the management of patients with bloodstream infections (BSIs) to facilitate optimal antibiotic therapy selection. Recent technological advances have resulted in several rapid methods for determining susceptibility direct from positive blood culture that can provide turnaround times in under 8 h, which is considerably shorter than conventional culture-based methods. As diagnostic results do not directly produce a medical intervention, actionability is a primary determinant of the effect these technologies have on antibiotic use and ultimately patient outcomes. Randomized controlled trials and observational studies consistently show that rapid AST significantly reduces time to results and improves antimicrobial therapy for patients with BSI across various methods, patient populations and organisms. To date, the clinical impact of rapid AST has been demonstrated in some observational studies, but randomized controlled trials have not been sufficiently powered to validate many of these findings. This article reviews various metrics that have been described in the literature to measure the impact of rapid AST on actionability, antibiotic exposure and patient outcomes, as well as highlighting how implementation and workflow processes can affect these metrics.

Introduction

Approximately 20%–30% of patients with bloodstream infections (BSIs) receive inadequate antibiotic therapy.^1–3 At the same time, 67% of patients receive broad-spectrum antibiotics for empirical treatment of BSIs, and ultimately less than half of them have a resistant organism isolated.³ There is a need to determine antimicrobial susceptibility results faster than conventional culture-based methods, which take 24–48 h.^4,5 Several rapid methods for determining susceptibility are now available on the market with many others under development or in clinical trials that can provide turnaround times in the range of 1–8 h, which is considerably shorter than conventional culture-based methods.^6–15 These methods can be divided into two categories: (i) tests that look at the direct effect of antibiotics on bacteria [i.e. phenotypic tests or antimicrobial susceptibility tests (AST)], and (ii) tests that detect nucleic acid sequences indicative of resistance genes and their expression (i.e. genotypic tests or antimicrobial resistance tests). For the purpose of this review, the focus will be on rapid phenotypic AST (rapid AST) that provides results within 8 h or less (Table 1).¹⁷ The approaches to rapid phenotypic AST have been recently reviewed by Banerjee and Humphries,¹⁸ as well as Vasala and colleagues,¹⁹ and will not be discussed in detail here. To date, evidence for rapid AST exists primarily for positive blood culture specimens and this will also be the focus of this article.

Table 1.

Commercially available rapid phenotypic antimicrobial susceptibility testing methods

Test	Reference	Sample type	Average time to results (h)	Regulatory status
Alfred 60/AST (Alifax®, Polverara, Italy)	Barnini S et al.9	Gram-negative BC Gram-positive BC	4–6	CE-IVD
ASTar® system (Q-linea, Uppsala, Sweden)	Göransson J et al.7	Gram-negative BC	6	US-FDA, CE-IVD
FASTinov (FASTinov SA, Porto, Portugal)	Costa-de-Oliveira S et al.15	Gram-negative BC Gram-positive BC	2	CE-IVD
LifeScale system (Affinity Biosensors, Santa Barbara, CA, USA)	Synder JW et al.16	Gram-negative BC	4.5	US FDA, CE-IVD
Next-Generation Phenotyping (NGP) system (Selux, Charlestown, MA, USA)	Baker et al.14	Gram-negative BC Gram-negative isolated colonies Gram-positive isolated colonies	5.5	US FDA, CE-IVD
Pheno® system (Accelerate Diagnostics, Tucson, AZ, USA)	Cenci et al.8	Gram-negative BC Gram-positive BC	7	US FDA, CE-IVD
QMAC-dRAST™ (QuantaMatrix, Inc., Seoul, Republic of Korea)	Kim et al.6	Gram-negative BC Gram-positive BC	6	CE-IVD, MDFS Korea
VITEK® REVEAL™ (bioMérieux, Marcy-l'Etoile, France)	Tibbetts et al.12	Gram-negative BC	5.5	US FDA, CE-IVD

BC, blood culture; CE-IVD, Conformité Européenne-in vitro diagnostic; MDFS, Ministry of Food and Drug Safety.

Objectives

This article will review and summarize various metrics that have been described in the literature to measure the impact of rapid AST on actionability, antibiotic exposure and patient outcomes. In addition, the article will highlight the importance of process and workflow optimization and their effect on the metrics used to evaluate the success of rapid AST implementation.

Impact of rapid AST on actionability and clinical decision making

As more rapid AST methods become accessible, the actionability of the test results (i.e. the ability of test results to influence clinical decision making) is often used to assess whether they provide any advantage over conventional phenotypic methods. In the context of this review, actionability shall be defined as the rapidity and/or frequency of changes in antibiotic treatment attributed to the test results. As diagnostic results do not directly produce a medical intervention, actionability is a primary determinant of the effect these technologies have on antibiotic use and ultimately patient outcomes. As patient outcomes may be impacted by a complication or condition beyond antimicrobial chemotherapy (e.g. the need for surgical intervention), evaluating the actionability may be a more direct measure to assess the value of rapid AST.

To be actionable, rapid AST results must provide sufficient information for clinicians to assess the appropriateness of the patient’s current antibiotic therapy and be able to determine which antibiotic regimen is optimal for the patient. Ideally, the antibiotic modifications based on rapid AST data should result in the patient receiving optimal antibiotic therapy sooner (where optimal could be defined as the narrowest effective antibiotic based on antimicrobial susceptibility results), without the need for further antibiotic modification or additional susceptibility testing. Rapid tests that report the MIC allow for optimization of antibiotic therapy in the individual patient based on antimicrobial pharmacokinetic/pharmacodynamic principles. However, data on the impact of this approach on the actionability of rapid AST in clinical practice are limited.²⁰

Due to the uncertainty of the causative pathogen at presentation, more than two-thirds of patients receive broad-spectrum antibiotics for empirical treatment of BSIs in an effort to minimize the risk of undertreatment.³ Overly broad therapy, or overtreatment, can provide selective pressure for emergence of antimicrobial-resistant bacteria, Clostridioides difficile infection and antibiotic-associated adverse events.^21–23 On the other hand, despite the common use of empirical broad-spectrum antibiotics, approximately one in five patients receives inadequate empirical antibiotics (i.e. failure to effectively treat the causative organism).¹ As both inadequate and unnecessarily broad empirical antibiotics have been associated with higher rates of mortality, and both are eligible for antibiotic modification to achieve optimal treatment, assessing the potential actionability of rapid AST results among these patient cohorts should be performed prior to adoption.³ There are two main potential advantages of rapid AST in terms of actionability, namely, rapidity and frequency. The speed with which antibiotic treatment can be modified is inherent to the faster turnaround time of the test, whereas increase in the frequency is in response to the increased effort, resources and attention placed on rapid AST results. It is worth noting that in some circumstances, rapid AST results will confirm that the empirical antibiotic prescribed is the optimal agent for the patient and no further antibiotic modifications are needed. Although the rapid AST results will likely reduce the time period of diagnostic uncertainty, it is difficult to measure and quantify the benefit of rapid results in these circumstances as there will be no associated actionability. For this reason, it is unlikely that rapid AST results will lead to an antibiotic modification in 100% of cases, but the rates of undertreatment and overtreatment with empirical prescribing suggest there is a large proportion of patients with opportunity for improved antibiotic use. Therefore, rapid AST can inform timely decisions on the choice of optimal ‘personalized (to the circulating isolates) therapy’. Whereas rapid AST tests are the enabler of actionability, implementation is dependent on the speed of reporting and response of clinical staff (discussed in the next section), which is beyond the control of the laboratory running the rapid tests.

How implementation and laboratory workflow impact the actionability and outcomes of rapid AST

Implementing diagnostic tests that improve patient care through expedited time to results is a core function of the clinical microbiology laboratory. Rapid AST tests have the potential for widespread adoption in clinical microbiology laboratories if they can be effectively implemented and shown to generate results equivalent to conventional AST methods. Doing so will require microbiology laboratories to identify and overcome hurdles to implementation, including potential adjustments in laboratory workflow to optimize testing and reporting so there is minimal delay in the potential actionability of results.

At a minimum, microbiology laboratories will need to validate the new technology against reference methods and ensure that all compliance and regulatory requirements are met. As with many new technologies, the uptake of rapid AST methods into clinical practice can be challenged by verification and quality process requirements. Many microbiology laboratories perform verification to confirm satisfactory performance of the new tests in their hands, to comply with standards and local recommendations, and to compare the new test performance against the existing method in use.^24–27 Verifications are usually to confirm the accuracy, reproducibility and quality control of the test, but must be adapted according to laboratory specifications and local regulations.

Beyond these regulatory and compliance hurdles, the setting in which rapid AST is deployed can impact the actionability and clinical utility, as it can significantly affect the time savings compared with existing test methods. There are several practical and financial aspects that must be considered. Because the lab will be the primary cost centre for the test, they will be asked to justify the costs of equipment and supplies, as well as ongoing maintenance expenses. The rapid pathogen identification test will need to accompany rapid AST implementation, if not included in the same test. Information flow is a major hurdle in operationalizing rapid AST. Tests without seamless integration to hospital information technology (IT) systems may not be adopted, as the process to integrate the test into the laboratory information system can be time consuming and labour intensive if the new test requires many manual process steps for compatibility and ‘talking’ across systems. Tests that have IT integration are more appealing and functional with respect to laboratory workflow as IT support is often complex and slow at many institutions. The compliance to the site’s cybersecurity requirements for the new systems is another key component in the implementation of rapid AST systems.

Laboratory working hours

Logistically speaking, the lab will need to provide staff training and ensure proficiency testing for the new technology. Larger academic medical centres are more often equipped with the infrastructure and resources needed to address these factors and represent the early adopters of rapid AST methods. The utility of rapid AST in smaller, non-academic centres can be limited by staffing models with trained personnel who are able to perform testing. The laboratory working shifts and availability of trained lab personnel to set up AST are key issues in the implementation of rapid AST. Many microbiology laboratories do not have adequately trained staff covering second and third shifts, which may directly impact the timely set-up and release of rapid AST results. A specimen set up for rapid AST on the first shift may yield results during the second shift when a trained technologist is unavailable to review and release results. Although this may be achievable at academic or large institutions that operate on a 24/7 basis, it may not be possible for institutions that are not open 24/7 or do not have staff available to perform rapid AST on second and third shifts. As a result, some laboratories will only be able to perform rapid AST during certain hours of the day, or they may perform rapid AST 24/7 but may not place overnight results into the electronic medical record (EMR) until they are verified by a medical laboratory scientist, which may not occur until the following morning. However, a retrospective modelling study in this Supplement found that the clinical value of rapid AST results can be effectively utilized even in centres that process positive blood cultures during a single shift.²⁸

Laboratory location and workflow processes

Microbiology laboratories that provide services for a multi-hospital system should optimize the time between specimen collection and specimen receipt in the lab prior to implementing rapid AST. Sending samples to a centralized location may take away some of the potential time savings due to the transport time alone. Although centralization or consolidation of microbiological services may enable more rapid diagnostic technologies to be offered, this must be balanced against logistical delays in receiving samples that could be mitigated by faster on-site testing.

As with many new laboratory instruments, rapid AST can have three main roles—replacement, targeted or add-on. The role under which rapid AST is implemented can affect the actionability and should be considered when assessing the impact. For instance, targeted testing of selected critically ill populations may skew the actionability towards antibiotic escalations, whereas targeting testing of stable patients may promote more antibiotic de-escalations and transitions to oral antibiotics. The laboratory workflow and staffing model, patient population and features of the test itself are key factors when defining the role of rapid AST. A few important questions to address when determining the role of rapid AST may include:

1. Does the test require dedicated staff to perform, or can it be integrated into existing workflow?

2. How will one identify patients and implement strategies for targeted testing?

3. Is the test cost-effective?

All of these implementation and workflow processes should be evaluated and optimized, to bridge any delays from sample collection to result, and to make sure that test results are being communicated effectively and in a timely manner to optimize the patient’s care.

How reporting and communication of results affect the actionability and outcomes of rapid AST

Like other infectious disease diagnostics, the actionability of rapid AST is likely to be affected by several factors beyond the test results. The impact of rapid AST on patient outcomes would be limited unless results are reported, communicated, and action is taken in timely fashion. For many institutions, a preliminary report of the positive blood culture is phoned as a critical lab result to the wards as soon as the Gram stain morphology of the organism is known, along with rapid molecular testing results, if performed. AST results, which are usually not available until 1–2 days later, are not routinely communicated as critical lab results. Laboratories may need to re-evaluate their reporting of positive blood cultures, as the availability of rapid AST results may warrant a modification to the result communication process and prioritization. Failure to notify the clinician in a timely manner may lead to lower actionability and decreased impact on patient outcomes.

Role of antimicrobial stewardship in the actionability and outcomes of rapid AST

For rapid infectious diseases diagnostics in the acute care setting, it is generally accepted that testing should be paired with antimicrobial stewardship (AMS) to maximize actionability and impact on antibiotic use and patients’ outcomes.²⁹ In fact, the joint IDSA and the Society for Healthcare Epidemiology of America guidelines for implementing an AMS programme recommend rapid diagnostic testing in addition to conventional culture and routine reporting on blood specimens if combined with active support and interpretation, but this recommendation is weak with moderate quality of evidence.³⁰

Methods used to communicate results to end users

Various approaches to direct communication of results (see ‘Active intervention’ in Table 2) for rapid AST have been reported in the literature, including real-time notification of results by text, page, telephone call or clinical decision support software (CDSS) alert. Recipients of the direct communication range from AMS providers, infectious diseases physicians, pharmacists and nurses to other healthcare providers. In the Antibiotic Research Leadership Group–supported randomized controlled trial (RCT) evaluating the clinical impact of rapid AST (Accelerate PhenoTest^™ BC Kit) for Gram-negative bacteraemia (RAPIDS-GN), which was performed at two US academic medical centres, AMS providers were notified by page at the time of every positive Gram stain, organism identification (ID) and AST.³¹ Another approach is to incorporate the communication of rapid AST results into the patient care pathway for the management of Gram-negative bacteraemia (GNB). A study by Walsh et al.³² evaluated the impact of a bundled AMS intervention using rapid AST on the outcome of GNB. These investigators assembled a multidisciplinary task force to create a bundled intervention that included: (i) rapid AST testing; (ii) a clinical decision making algorithm; (iii) real-time communication of results to the AMS programme; and (iv) a prospective audit with real-time intervention and feedback.³² Educational lectures and presentations on the bundle were given to medical and house staff as part of the roll-out. These investigators found that the implementation of a bundled intervention improved the management of GNB without adversely affecting patient outcomes at their institution. In more resource-limited settings, the direct communication of rapid AST results could be limited to selected organisms or resistance phenotypes that are likely to require antibiotic modification. For example, Sheth et al.³³ reported improvements in patient care when implementing rapid AST with the real-time notification of targeted organisms (Pseudomonas aeruginosa, Acinetobacter baumannii or Candida spp.) deemed to be ‘critical results’, in an effort to limit the number of communications, particularly those occurring on second and third shifts.

Table 2.

Studies reporting process measures related to actionability of rapid AST for BSI

Test	Reference	Infection type	Study design	Laboratory testing/reporting hours	Active intervention	Antibiotic modification	Escalation	De-escalation
Alfred 60/AST (Alifax®, Polverara, Italy)	Anton-Vazquez, 2022, UK58	Gram-negative BSI (n = 191)	Prospective, single-centre, service evaluation	Performed before noon (weekdays only)	Electronically reported or actively communicated by telephone by clinical microbiologist to attending clinician	Decreased TOT No difference in TET	Faster escalation	NR
EUCAST RAST	Cardot Martin, 2022, France73	E. coli, K. pneumoniae, P. aeruginosa BSI (n = 110)	Prospective, single-centre, cohort	Performed in the morning (weekdays only) Reported in the afternoon (weekdays only)	Daily monitoring of positive blood cultures	Increased AET No difference in TET	NR	NR
	Berinson, 2021, Germany74	E. coli, K. pneumoniae, P. aeruginosa, A. baumannii complex BSI (n = 105)	Pre-/post-, quasi-experimental, single-centre	Performed in the morning (7 d/wk)	RTN by the microbiologist to the clinician	Decreased TOT No difference in AOT	NR	NR
Mueller–Hinton Rapid agar (i2a, Montpellier, France)	Pilmis, 2019, France75	Enterobacteriaceae or S. aureus (n = 330)	Pre-/post-, quasi-experimental, single-centre	Performed during business hours (7 d/wk)	RTN by telephone to AMS	Faster antibiotic modification	No difference in % escalations	More de-escalations
PhenoTest® BC Kit (Accelerate Diagnostics, Tucson, AZ, USA)	Banerjee, 2021, USA31	Gram-negative BSI (n = 448)	Prospective, multicentre, RCT	Performed weekdays only during local laboratory business hours (24/7 at site 1 but site 2 was closed from 2.00 am to 6.00 am) Result reported 24/7	RTN via page for each GS, ID and AST result (business hours during weekdays only)	More antibiotic changes Faster antibiotic modification	Faster escalation	No difference in time to de-escalation
	Bhalodi, 2022, USA61	Gram-positive and Gram-negative BSI (n = 854)	Pre-/post-, quasi-experimental, multicentre	Performed 24/7 (overnight results reported following am at 1 site)	RTN at 3 sites BC list audit and feedback at 2 sites (weekdays only)	Faster antibiotic modification Decreased TOT Increased AOT	No difference in time to escalation	Decreased time to de-escalation
	Robinson, 2021, USA59	Adult, Gram-negative BSI (n = 514)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7 Results reported business hours 7 d/wk	RTN via email (business hours during weekdays only)	Decreased TOT No difference in AOT	NR	NR
	Ehren, 2020, Germany76	Gram-positive cocci in clusters, Gram-positive cocci in short chains, Gram-negative rods BSI (n = 204)	Pre-/post-, quasi-experimental, single-centre	Performed during business hours during weekdays only	RTN with AMS chart review and bedside consultation (business hours during weekdays only)	No difference in TET Decreased TOT Increased AOT	No difference in time to escalation	Decreased time to de-escalation
	Walsh, 2021, USA32	Adult, Gram-negative BSI (n = 206)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7	RTN to the ordering clinician via phone call as well as to the AMS via text page/CDSS	No difference in TET Decreased TOT Decreased time to oral therapy	Faster escalation	Decreased time to de-escalation
	Sheth, 2020, USA33	Adult, Gram-negative BSI (n = 173)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7 on weekdays (business hours on weekends)	RTN 24/7 of ‘critical results’ (P. aeruginosa, A. baumannii, or Candida spp.) Audit and feedback (business hours during weekdays only)	Faster antibiotic modification Decreased TOT	NR	More de-escalations
	MacVane, 2021, USA60	Gram-positive BSI (n = 219)	Pre-/post-, quasi-experimental, multicentre	Performed 24/7 (overnight results reported following am at 1 site)	RTN at 1 site (business hours during weekdays only) BC list audit and feedback at other site (business hours during weekdays only)	Faster antibiotic modification Decreased TOT	NR	NR
	Babowicz, 2021, USA67	Adult, Gram-negative BSI with concurrent sepsis (n = 116)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7 Results reported business hours 7 d/wk	Audit and feedback (business hours during weekdays only)	No difference in TET	NR	NR
	Brosh-Nissimov, 2023, Israel77	Gram-negative BSI (n = 103)	Pre-/post-, quasi-experimental, single-centre	Performed during business hours	Push notifications to alert patients’ department and the infectious diseases consultant after GS of a PBC bottle, and upon ID of GNB	No difference in TET Decreased TOT	NR	NR
	Ganapathiraju, 2022, USA78	Adult, Enterobacterales BSI (n = 255)	Retrospective, single-centre, cohort	NR	Daily monitoring of positive blood cultures (business hours during weekdays only)	NR	No difference in % escalation	No difference in % de-escalation
	Truong, 2022, USA79	Paediatric, Gram-negative BSI (n = 162)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7	Audit and feedback (business hours during weekdays only) + RTN via CDSS	Decreased TOT	No difference in % escalation	No difference in % de-escalation
	Dare, 2021, USA38	Gram-positive and Gram-negative BSI (n = 496)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7 (overnight results reported following am)	Daily monitoring of positive blood cultures (business hours during weekdays only) ± RTN (business hours 7 d/wk)	Decreased TOT Increased AOT	NR	NR
QMAC-dRAST (QuantaMatrix, Inc., Seoul, Republic of Korea)	Kim, 2021, Republic of Korea68	Adult (aged ≥16 y) patients with haematological malignancy (n = 116)	Prospective, single-centre RCT	Performed during business hours (weekdays only)	RTN to infectious diseases physicians by text message	Decreased TOT Increased AOT	NR	NR
Vitek AutoMicrobic system (Vitek Systems, Inc., Hazelwood, MO, USA)a	Trenholme, 1989, USA66	Gram-positive and Gram-negative BSI (n = 226)	Prospective, single-centre, RCT	NR	Audit and feedback	More antibiotic changes	More escalations	NR
Vitek 2 (bioMérieux, Marcy l’Étoile, France)a	Hogan, 2020, USA36	Adult, Enterobacterales or P. aeruginosa BSI (n = 671)	Pre-/post-, quasi-experimental, single-centre	Performed 24/7 Results reported 24/7 with templated comments	RTN via email (business hours during weekdays only)	No difference in time to oral therapy	No difference in % escalation No difference in time to escalation	No difference in % de-escalation No difference in time to de-escalation
	Christensen, 2022, USA35	Adult Gram-negative BSI (n = 205)	Prospective, single-centre, RCT	Performed 24/7	RTN via page (business hours during weekdays only)	No difference in TOT Increased use of oral therapy Decreased time to oral therapy	NR	NR

AET, achievement of effective antibiotic therapy; AMS, antimicrobial stewardship; AOT, achievement of optimal antibiotic therapy; AST, antimicrobial susceptibility testing; BC, blood culture; BSI, bloodstream infection; CDSS, clinical decision support software; GNB, Gram-negative bacteraemia; GS, Gram stain; ID, identification; NR, not reported; PBC, positive blood culture; RCT, randomized controlled trial, RTN, real-time notification; TET, time to effective therapy; TOT, time to optimal therapy.

^aOff-label use of direct inoculation with a sample from a positive blood culture vial.

AMS working hours

Realization of the maximum improvement in microbiology process measures, such as time to results, may require a 24/7 laboratory staffing model so that testing of blood culture specimens can be performed as soon as they signal positive. However, the impact of 24/7 testing on clinical decision making and patient outcomes is less clear. A study evaluating factors that impact how rapid genotypic test results are acted upon in a children’s hospital found that time to action was slower for results that were reported at night than during the day, which might have been related to the availability of the AMS team or that results that occur on third shift are likely not acted on until the following morning rounds.³⁴ A randomized trial of a laboratory-developed rapid AST method observed a change in the impact of rapid AST on the primary outcome of the study when assessing the working hours of AMS.³⁵ Time to narrowest antibiotic therapy was significantly reduced within the rapid AST group during AMS working hours [median (IQR), 93 (56–154) h versus 62 (43–91) h; P = 0.004], but not during off-hours [median (IQR), 73 (60–138) h versus 76 (52–115) h; P = 0.56]. In another rapid AST study, Hogan et al.³⁶ found that when restricted to cases with an accepted AMS intervention, median time to appropriate escalation or de-escalation in the rapid AST group was reduced compared with that in the rapid AST group as a whole (which included patients without AMS intervention), with median times of 52.3 versus 38.4 h. This is consistent with the findings from previous studies with rapid diagnostic tests that response to rapid testing is slower outside working hours.^29,37 Taken together, these studies suggest that in terms of actionability, 24/7 testing may not affect antibiotic use metrics around the clock unless the result can be reported, communicated and used to influence medical decision making in a timely manner.

Impact of direct communication on patient outcomes

In principle, real-time direct communication of rapid AST results should enable immediate antibiotic management decisions. Whereas some studies report that the use of real-time notification provides additional benefits for rapid diagnostic tests, others have shown mixed results (Table 2). Dare et al.³⁸ evaluated the impact of rapid AST with two different methods of AMS prospective audit and feedback, one with real-time notification (RTN) of results and the other without RTN. During the real-time notification intervention period (rapid AST + RTN), laboratory personnel would call rapid AST results to an AMS provider from 9 am to 5 pm, 7 days per week, with overnight batched results called each morning. The AMS provider would perform interventions at the time of notification as appropriate. During the other intervention period, laboratory personnel would report rapid AST results in the EMR per laboratory workflow without RTN, and the AMS team would review results during review of a positive blood culture list Monday through Friday and perform interventions as appropriate. There were improvements in subsets of antimicrobial use in the rapid AST + RTN that were not observed in the rapid AST without RTN period. Patients in the rapid AST + RTN period received more narrow β-lactams overall, and less vancomycin for patients with likely contaminated blood cultures, suggesting that RTN has potential added benefit in some scenarios. However, there were several laboratory and clinical outcomes that were significantly improved with or without RTN, compared with a historical cohort without rapid AST, suggesting that rapid AST even without RTN can have benefit (improved antibiotic process measures and shorter length of stay). Similar to Dare et al., it is worth mentioning that most studies that have used real-time direct communication of rapid AST results have done so during business hours and/or during weekdays only, further supporting that 24/7 testing and reporting may not be required to achieve meaningful improvements in patient care.

It is important for institutions to evaluate pros and cons of each communication method for their individual AMS programme resources and priorities. This must be balanced against staffing requirements not typically available outside of academic or large institutions. Ultimately, further studies that compare clinical outcomes and economic impacts with various implementation and notification strategies are needed to help determine where each intervention mechanism best fits into clinical practice.

Actionability, antibiotic use and patient outcome metrics: a review of existing evidence

Although the need for and benefit of providing more rapid AST results may seem self-evident, it is important to demonstrate the promised medical value given that the cost and resources required to implement and routinely perform these tests are significant. Many rapid AST studies have been published on performance of the assay and time savings, demonstrating acceptable performance and significantly reducing the time to results.^{7–9,39–55} Outside of a few multicentre studies, most of the clinical evidence comes from single-centre studies that were performed as part of their internal programme decision making, and the selection of metrics that were tracked may have been tailored to the individual institutional needs. As a result, tracking and reporting of metrics among rapid AST studies performed to date have been heterogeneous and have not comprehensively measured downstream impact on actionability, antibiotic use, patient outcomes or healthcare costs, which is a barrier to summarizing the literature and comparing different approaches.

Impact of rapid AST on actionability

Most data come from single health system pre-post quasi-experimental studies conducted in the USA or Europe, evaluating Gram-negative BSIs, with some element of AMS intervention present in each study arm. Table 2 summarizes selected studies and reviews the actionability of rapid AST in terms of time-to-event and/or the proportion of patients with (i) antibiotic modifications, (ii) antibiotic escalation, and (iii) antibiotic de-escalation. Comparison of actionability measures between these studies is limited by the variability in rapid AST method, patient population and organisms studied.

Antibiotic modifications

The underlying goal of rapid AST is to improve antibiotic use. Frequency of antibiotic modifications, or time to first antibiotic modification, is a process measure that does not directly measure appropriateness of use but is a straightforward metric that is routinely reported to evaluate the actionability of rapid AST. Many studies report faster antibiotic modifications with rapid AST, including the RAPIDS-GN RCT publication.³¹ In this study, the time from randomization to first antibiotic change was faster in the rapid AST arm than in the conventional AST arm [median (IQR), 8.6 (2.6–27.6) versus 14.9 (3.3–41.1) h; difference, 6.3 h; P = 0.02]. This study also reported an increased frequency of Gram-negative antibiotic changes due to rapid AST (55% versus 33% with conventional AST) during the initial 24 h following Gram stain.

Effective antibiotic therapy

Time to effective (or appropriate) therapy is an important metric that is most associated with clinical outcomes such as survival.^56,57 However, because empirical therapy is administered prior to rapid AST testing, a high proportion of patients will achieve effective therapy prior to the studied intervention. Therefore, it is the opinion of the authors that this metric may be most informative in rapid AST studies when assessed in the subgroup of patients who did not receive effective therapy at the time of the intervention (i.e. time to effective therapy is >0 h). In studies that have applied this approach, the impact of rapid AST on time to effective therapy has been significant.^31,32,36,58 For example, Walsh et al.³² found that in the subgroup of patients not on effective therapy at time of Gram stain, median time to effective therapy improved from 51.2 to 11.2 h (P < 0.001) with rapid AST, whereas the overall population had similar times with and without rapid AST. A similar relationship can be observed when evaluating the frequency of antibiotic escalation. Hogan et al.³⁶ did not observe any difference in the frequency of antibiotic escalation in their overall study population (36.4% versus 35.1%, P = 0.7), but when restricting to a subgroup of patients with resistant pathogens, antibiotic escalation was more frequent in the rapid AST group (54.0% versus 30.2%; P = 0.0001). In the RAPIDS-GN study, time to antibiotic escalation was shorter for antimicrobial-resistant BSIs [median (IQR), 18.4 (5.8–72.0) versus 61.7 (30.4–72.0) h; P = 0.01], but not in the antimicrobial-susceptible BSI population [median (IQR), 72.0 (6.2–72.0) versus 72.0 (24.2–72.0)].³¹

Optimal antibiotic therapy

Optimal therapy is one of the more commonly reported antibiotic process measures related to actionability that incorporates a quality of antimicrobial use into the metric. Determination of optimal therapy is more difficult to outline than time to effective therapy as it is often more subjective and there is no universal definition. One consideration is that optimal therapy is the antimicrobial drug(s) administered for the patient’s infection that is considered the institution’s most preferred treatment option for the patient based on the susceptibility profile, site of infection, patient’s condition and comorbidities, and hospital policy.^59–61 Based on that definition, it is likely that optimal therapy regimens will differ from institution to institution, making direct comparisons between studies unreliable. Nevertheless, most rapid AST studies have observed improvements in the time to optimal therapy (and achievement of optimal therapy) with rapid AST using various definitions, patient populations and organisms.

De-escalation

Whereas modification of antibiotic therapy has a clear medical value for patients receiving inadequate empirical antibiotics (i.e. failure to cover organisms), support for rapid antibiotic de-escalation for patients receiving adequate empirical antibiotics is limited. The concept is founded on the ecological risk associated with high levels of broad-spectrum antimicrobial use, but evidence remains inconclusive.^22,62–64 Nevertheless, de-escalation from broad-spectrum to narrow-spectrum antimicrobial therapy is an accepted, basic strategy of AMS that is advised in many international guidelines.^30,65 Moreover, early de-escalation of broad-spectrum therapy appears to be safe and may lower the risk of unintended consequences associated with broad-spectrum exposure.²¹ Despite the potential limitations of de-escalation as an outcome measure, it has been shown to be responsive to rapid AST and remains a frequently reported metric (Table 2).

Intervention rate

AMS interventions, as well as their acceptance rate, are process metrics that can be documented to capture the type and frequency of recommendations that are being made based on rapid AST results. Reporting of these interventions serves multiple purposes. First, it would provide objective data to quantify the number of interventions made and patients impacted. This helps exemplify and provides justification to expand resources when appropriate. Second, data on acceptance rates can be used to understand trends that an institution can use to optimize their implementation of rapid AST, as well as identify educational opportunities for quality improvement. Third, characterization of the type of intervention made would help to inform whether targeted testing would be beneficial or if the current testing process is optimal. Notably, in two RCTs (RAPIDS-GN³¹ and Trenholme et al.⁶⁶), AMS made more recommendations in the rapid AST group than in the conventional AST group. Sheth et al.³³ reported a higher percentage of AMS recommendations (40.4% versus 19.0%, P = 0.002) in the rapid AST group compared with the pre-group, which used a rapid genotypic test.

Outcomes

Table 3 summarizes the published studies that have reported outcomes metrics associated with rapid AST for BSI. The most commonly used metrics measure antimicrobial exposure, clinical outcomes and length of stay.

Table 3.

Studies reporting outcome measures on rapid AST for BSI

Test	Reference	Infection type	Study design	Antibiotic exposure outcomes	Clinical outcomes	Length of stay and costs
16S rRNA PCR (growth in presence of antibiotics)	Beuving, 2015, the Netherlands80	Adult, Gram-positive cocci and Gram-negative bacilli BSI (n = 152)	Pre-/post-, quasi-experimental, single-centre	NR	No difference in mortality	No difference in LOS
Alfred 60/AST (Alifax®, Polverara, Italy)	Anton-Vazquez, 2022, UK58	Gram-negative BSI (n = 191)	Prospective, single-centre, service evaluation	Less aminoglycosides	No difference in mortality	No different in LOS
EUCAST RAST	Cardot Martin, 2022, France73	E. coli, K. pneumoniae, P. aeruginosa BSI (n = 110)	Prospective, single-centre, cohort	NR	No difference in mortality	No difference in LOS
	Valentin, 2021, Austria81	Gram-positive cocci and Gram-negative bacilli BSI (n = 152)	Pre-/post-, quasi-experimental, single-centre	NR	No difference in mortality	NR
	Berinson, 2021, Germany74	E. coli, K. pneumoniae, P. aeruginosa, A. baumannii complex BSI (n = −105)	Pre-/post-, quasi-experimental, single-centre	NR	No difference in mortality	NR
PhenoTest® BC Kit (Accelerate Diagnostics, Tucson, AZ, USA)	Banerjee, 2021, USA31	Gram-negative BSI (n = 448)	Prospective, multicentre, RCT	No difference in duration of broad-spectrum antibiotics	No difference in mortality, CDI, MDRO, readmission	No difference in LOS, total or direct costs of hospitalization
	Bhalodi, 2022, USA61	Gram-positive and Gram-negative BSI (n = 854)	Pre-/post-, quasi-experimental, multicentre	NR	No difference in mortality, AKI, CDI, MDRO, readmission	No difference in LOS
	Robinson, 2021, USA59	Adult, Gram-negative BSI (n = 514)	Pre-/post-, quasi-experimental, single-centre	Less broad-spectrum antibiotics More narrow-spectrum β-lactams	No difference in mortality, CDI, readmission, BSI relapse	No difference in LOS
	Ehren, 2020, Germany76	Gram-positive cocci in clusters, Gram-positive cocci in short chains, Gram-negative rods BSI (n = 204)	Pre-/post-, quasi-experimental, single-centre	No difference in duration of therapy or DOT	No difference in mortality, CDI	No difference in LOS
	Walsh, 2021, USA32	Adult, Gram-negative BSI (n = 206)	Pre-/post-, quasi-experimental, single-centre	Decreased duration of therapy	No difference in readmission, BSI relapse	Shorter LOS
	Sheth, 2020, USA33	Adult, Gram-negative BSI (n = 173)	Pre-/post-, quasi-experimental, single-centre	Decreased antibiotic intensity score Decreased duration of broad-spectrum antibiotics	No difference in mortality, readmission	Shorter LOS
	MacVane, 2021, USA60	Gram-positive BSI (n = 219)	Pre-/post-, quasi-experimental, multicentre	Decreased duration of broad-spectrum antibiotics	No difference in mortality, AKI, CDI, MDRO, readmission	No difference in LOS
	Babowicz, 2021, USA67	Adult, Gram-negative BSI with concurrent sepsis (n = 116)	Pre-/post-, quasi-experimental, single-centre	Increased narrow-spectrum Gram-negative antibiotic duration of therapy Decreased antipseudomonal antibiotic duration of therapy	Decreased mortality Increased sepsis resolution	No difference in LOS
	Brosh-Nissimov, 2023, Israel77	Gram-negative BSI (n = 103)	Pre-/post-, quasi-experimental, single-centre	Increased duration of therapy	No difference in mortality, CDI, readmission	No difference in LOS
	Ganapathiraju, 2022, USA78	Adult, Enterobacterales BSI (n = 255)	Retrospective, single-centre, cohort	NR	No difference in mortality	No difference in LOS
	Truong, 2022, USA79	Paediatric, Gram-negative BSI (n = 162)	Pre-/post-, quasi-experimental, single-centre	Decreased meropenem duration of therapy	No difference in mortality	No difference in LOS
	Dare, 2021, USA38	Gram-positive and Gram-negative BSI (n = 496)	Pre-/post-, quasi-experimental, single-centre	Decreased overall and broad-spectrum Gram-negative antibiotic DOT	No difference in mortality, CDI	Shorter LOS
	Roth, 2022, USA69	Gram-positive and Gram-negative BSI (n = 290)	Pre-/post-, quasi-experimental, single-centre	NR	No difference in mortality, CDI	Shorter LOS
QMAC-dRAST (QuantaMatrix, Inc., Seoul, Republic of Korea)	Kim, 2021, Republic of Korea68	Adult (aged ≥16 y) patients with haematological malignancy (n = 116)	Prospective, single-centre RCT	Less broad-spectrum antibiotics	No difference in mortality, CDI, MDRO, time to defervescence	NR
Vitek 2 (bioMérieux, Marcy l’Étoile, France)a	Hogan, 2020, USA36	Adult, Enterobacterales or P. aeruginosa BSI (n = 671)	Pre-/post-, quasi-experimental, single-centre	NR	No difference in mortality, AKI, CDI	No difference in LOS
	Christensen, 2022, USA35	Adult Gram negative BSI (n = 205)	Pre-/post-, quasi-experimental, single-centre	No difference in duration of therapy or DOT	No difference in mortality, readmission, BSI relapse	Shorter LOS

AKI, acute kidney injury; AST, antimicrobial susceptibility testing; BSI, bloodstream infection; CDI, Clostridoides difficile infection; DOT, days of therapy; LOS, length of stay; MDRO, multidrug resistant organism; NR, not reported; RCT, randomized controlled trial.

^aOff-label use of direct inoculation with a sample from a positive blood culture vial.

Antibiotic exposure outcomes

Although the driving force for rapid AST implementation for institutions may be to improve patient outcomes and reduce healthcare costs, measures of antibiotic usage are a core element of AMS interventions that should be responsive to improved actionability from faster AST results. Many of the antibiotic usage and outcome measures that have been used to evaluate rapid genotypic tests are also used to measure the impact of rapid AST. However, there is no clear, validated, standard definition for these metrics. In general, rapid AST improved antibiotic usage for a variety of measures including less broad-spectrum antibiotic consumption through improved use of anti-MRSA and anti-pseudomonal agents, more use of narrow-spectrum agents, decreased antibiotic intensity score, and decreased duration of therapy (Table 3).^{32,33,59,60,67,68}

Clinical outcomes

A recent Cochrane review concluded that the theoretical benefits of rapid AST have not been demonstrated in randomized studies to directly improve mortality or length of stay.¹⁷ Moreover, the rate of MDR Gram-negative infections among patients enrolled in rapid AST RCTs has been low, which may contribute to the lack of effect on clinical outcomes. Yet, many institutions express a strong desire to improve these endpoints when evaluating new technologies. Clinical outcomes are often perceived as more impactful compared with antibiotic modifications and other process measures. However, it is important to remember that many clinical outcomes are multifaceted, with change difficult to detect, and may not be directly responsive to improvements in antibiotic prescribing. Moreover, there is little guidance and a lack of consensus on how patient outcome metrics should be assessed (e.g. time-to-event analysis, use of subset populations, starting point for calculations, etc.). Risk adjustment for confounding factors (e.g. patient comorbidities, severity of illness, and rates of antimicrobial resistance) are infrequently performed, which could influence the clinical outcome findings in many of these studies.

The impact of rapid AST on patient outcomes other than mortality has been reported in only a few RCTs, which likely did not have large enough numbers to detect a difference in the most important outcomes.^17,31,68 With the exception of one study, observational studies of rapid AST have not reported significant differences in mortality, or other relevant endpoints related to antibiotic-associated adverse events, readmission or relapse. Babowicz et al.⁶⁷ reported a higher proportion of sepsis resolution and decreased risk of 30 day inpatient mortality in a pre-/post- quasi-experimental, single-centre study of 116 adult patients with Gram-negative BSI with concurrent sepsis.

Like mortality assessments, quantifying the impact of rapid AST on many of the other commonly reported clinical endpoints in studies is challenging. Generally, the incidence of these events is relatively low and therefore requires a large sample size to measure a true effect. Additionally, the fluctuations in these rates are impacted by more than just the receipt of antimicrobial therapy. For example, infection control measures can have a tremendous impact on the control of C. difficile infection rates. Some authors have proposed use of clinical outcomes as a complementary metric to reassure providers that interventions did not cause unintended negative clinical consequences.³² For example, hospital readmission and infection relapse may be clinical outcomes that are more informative to determine if ‘no harm’ came from more rapid clinical interventions facilitated by rapid AST.

Length of stay

Implementation of rapid AST has been associated with reductions in length of stay in many observational studies, particularly in selected infection types such as Gram-negative BSI.^{32,33,38,61,69} This could be due to many reasons including improvements in antibiotic process measures. For example, faster AST reporting along with AMS intervention can facilitate early transitions to oral therapy and hospital discharge.³⁵ Additionally, if the time to appropriate antibiotic therapy is reduced, a decrease in the patient’s infection-related length of stay can be envisioned.⁷⁰

RCTs and a small number of observational studies have not observed the length-of-stay benefit reported in many of the observational studies. Length of stay can be influenced by factors beyond antibiotic decision making, including patient comorbidities and severity of illness, requiring longer hospitalization to manage concurrent medical issues.

Whatever outcome is measured to demonstrate the actionability of the rapid AST, it should be kept in mind that rapid AST will need to be accompanied with streamlined workflow, quality AMS and rapid reporting to measure the actionable outcome. Further, it should be kept in mind that the study design, enrollment size and type of analysis will provide the required power for these studies in assessing the clinical decision making as an outcome of rapid AST implementation. For example, as an alternative to assessing the percentage change of an outcome parameter (such as mortality rate) in a pre-post implementation study, the desirability of outcome ranking in clinical trials might be more informative in impact assessments of rapid AST.^71,72

Conclusion

Through the implementation of new technologies, hospitals and healthcare professionals aim to improve the care of patients. RCTs and observational studies consistently support that rapid AST significantly reduces time to results and improves antimicrobial therapy for patients with BSI across various methods, patient populations and organisms. However, because healthcare resources are finite, new interventions must demonstrate cost-effectiveness or cost-benefit to be sustainable and provide benefit to all relevant stakeholders. For many institutions, this requires demonstrations of clinical impact through improvements in patient and/or economic outcomes. To date, the clinical impact of rapid AST has been demonstrated in some observational studies but RCTs have not been sufficiently powered to validate many of these findings. Consequently, adequately powered prospective studies should focus on reporting the most clinically meaningful outcomes.¹⁷ Validation and standardization of metrics should be another focus for future research, including consensus on whether evaluation of clinical outcomes should prioritize analyses based on the number of participants who are on effective or ineffective antibiotic therapy before susceptibility testing results are available. Future studies should be designed with endpoints that can inform implementation practices. Identifying what samples should be tested and how results can best be communicated to ensure rapid actionability would be beneficial to patients and healthcare providers.

Funding

This paper was published as part of a supplement financially supported by bioMérieux.

Transparency declarations

S.H.M. and H.P.D. are employees of bioMérieux.