Ethics statement
This was a single-center study conducted at Northwestern Memorial Hospital, an 894-bed academic medical center located in Chicago, Illinois. Study samples were residual clinical specimens not specifically collected for study purposes. The study was approved by the Institutional Review Board (IRB) at Northwestern University.
Study samples and microbiology laboratory process
Blood culture was performed with the BacT/ALERT® system (bioMérieuxbioMérieux, Durham, NC). Each blood culture included one aerobic bottle and one anaerobic bottle which were incubated for up to five days until positivity. Only one blood culture from each patient was used in the study. Positive blood cultures were processed for ID and AST 24 h a day and 7 days a week. Eighty-two non-duplicate patient samples were included in the study. In addition, 18 challenge isolates with well-characterized resistance mechanisms were also used. The challenging isolates were diluted to 1 × 106 CFU/ml with sterile saline and mixed with 10 ml of whole blood before inoculating into the blood culture bottle.
Once a blood culture flagged positive, Gram stain was performed followed by testing with the Accelerate PhenoTest™ BC kit within 8 h of growth detection. Only blood cultures positive for gram-negative bacilli were included in the study. The gram-negative testing panel includes Pseudomonas aeruginosa, Acinetobacter baumannii, Klebsiella spp. (i.e., Klebsiella pneumoniae, Klebsiella oxytoca, not differentiated), Escherichia coli, Enterobacter spp. (i.e., Enterobacter cloacae, Enterobacter aerogenes, not differentiated), Proteus spp. (i.e., Proteus mirabilis, Proteus vulgaris, not differentiated), Citrobacter spp. (i.e., Citrobacter freundii, Citrobacter koseri, not differentiated), and Serratia marcescens. Cultures positive for gram-negative rods were sub-cultured to a sheep’s blood agar plate (BAP), a chocolate agar plate (CHOC), and a MacConkey agar plate (MAC). For each sample, times were recorded for Gram stain notification, the AXDX ID and AST results, and the VITEK® MS ID and VITEK® 2 AST results as standard of care (SOC) via the electronic health record (EHR). VITEK MS ID was performed three times during the first shift, and once each during second and third shifts whenever sufficient growth was detect on plate. AST with the VITEK® 2 GN70 was performed with overnight plate growth. Cultures with more than one organism were sub-cultured before VITEK 2 testing. AST was performed once a day during the first shift. Aliquots of blood culture samples were frozen for later adjudication testing. AXDX was performed right after the Gram stain result was obtained.
Discrepancy resolution
Culture was repeated with the frozen samples that showed identification discrepancies between AXDX and SOC. Frozen samples were plated on BAP, CHOC, and MAC plates, and incubated at 35 °C for 24 h. The frozen aliquot was also sent for third-party testing with the VITEK® 2 system. When third-party testing confirmed an AXDX positive result or an AXDX negative result, the AXDX result was reported as a true positive or true negative result, respectively. When third-party testing did not confirm an AXDX positive result or an AXDX negative result, the AXDX result was reported as a false positive or false negative result, respectively.
The frozen isolates recovered from samples with discrepant AST categorical results were tested in triplicate with standard broth microdilution (BMD) from Clinical Laboratory Standards Institute (CLSI) to obtain a modal MIC result. Frozen isolates were sub-cultured twice before AST testing.
Stewardship simulation
Twenty-seven adult patients (≥ 18 years old) with ≥1 positive blood culture containing a gram-negative rod isolate were enrolled from February 2017 to May 2017. Informed consent was obtained per IRB protocol.
Routine SOC and AXDX were performed following the process described above. Our current institutional laboratory protocol dictates that a physician from the primary treatment team must be notified with the Gram stain result via phone in less than 1 h after bacterial growth is detected by BacT/ALERT®.
A Simulated ASTEW intervention was performed after AST results for the both VITEK® 2 and AXDX were available. A 2-h response time was allotted for a stewardship intervention if the antimicrobials needed to be optimized and the stewardship team was onsite. This 2-h turn-around time corresponds to the time needed for a stewardship team member to review the patient chart, contact the primary team, have the order processed by the pharmacy, and have the new antimicrobial delivered to the patient’s nurse for administration.
Given that the results were reported 24 h per day, varying degrees of simulated stewardship coverage were explored to determine possible impacts on times to ASTEW; these included shift times of 8-h (0800–1600), 16-h (0800–0000), and 24-h coverage. If the AXDX results were reported during off-hours, then the time to ASTEW was recorded 2 h after the start of the next active shift. For example, if the AXDX results were recorded at 0200 for simulated 16-h stewardship coverage, then the time to antimicrobial optimization based on ASTEW would be recorded at 1000 (0800 + 2 h).
As the control, antimicrobial stewardship team notification is not routinely performed but team members are available on-site during daytime hours (0800–1630) Monday through Friday and via pager off-site all other times to answer questions and make recommendations when necessary. AXDX results were not available to the stewardship team.
Antimicrobial optimization
Data on both empiric and targeted antimicrobial therapies were recorded. An antimicrobial therapy was defined as optimal when it was the narrowest spectrum agent with acceptable activity against the isolated pathogen based on the AST results. Every case was reviewed by a panel of 3 infectious disease specialists from the ASTEW team (2 infectious disease pharmacists and 1 infectious disease physician) and 2 of the 3 specialists had to agree for the agent to be considered optimal therapy. The time to optimal antimicrobial therapy was recorded from the time of Gram-stain notification to the time that the first dose of optimal therapy was administered as noted on the electronic medical record. If the antimicrobial therapy was not optimal, the time to optimal therapy was recorded as the time the patient was discharged or when the antimicrobial regimen was completed, whichever was sooner.
Outcomes evaluation
The primary outcome measure was the simulated difference in time to antimicrobial optimization when utilizing AXDX with 8-h ASTEW coverage compared to our institutional SOC without ASTEW support. If antimicrobial optimization occurred prior to AXDX AST results, the time difference was recorded as zero. Secondary outcome measures included the simulated differences in time to antimicrobial optimization when utilizing AXDX with 8-h, 16-h, and 24-h ASTEW coverage compared to our institutional SOC with 8-h ASTEW coverage.
Statistical analysis
Positive percent agreement (PPA) and negative percent agreement (NPA) were calculated for each AXDX ID result compared to the SOC result. In addition, positive predictive value (PPV) was calculated for the AXDX monomicrobial calls compared to SOC.
Essential agreement (EA), categorical agreement (CA), very major error (VME), major error (ME), and minor error (mE) rates were calculated as a way to measure the AST accuracy of AXDX compared to those of the SOC for each antimicrobial tested. EA is the percentage of the total test results within one doubling dilution of the SOC result. CA is the percentage of the total test results with the same categorical interpretation result as the SOC result. VME is the percentage of the resistant isolates by the SOC that tested susceptible by the AXDX. ME is the percentage of the susceptible isolates by the SOC that tested resistant by the AXDX. mE is the percentage of the total test results in which one result (from the AXDX or the SOC) is intermediate and the other is not.
The time difference to antimicrobial optimization versus the SOC arms were analyzed by paired student t-tests. A one-way ANOVA was utilized to assess a difference in optimization time between the three arms of 8, 16, and 24-h stewardship coverage. Statistical analyses were performed using Microsoft Excel 2013 and SPSS Version 23.0 (IBM Corp., Armonk, NY).