Study design

This single-center, retrospective, observational, before-and-after cohort study was conducted at a 625-bed academic medical center. Data were collected via electronic medical record (EMR) chart reviews of inpatient encounters from June 2018 to November 2018 (historical cohort) and from June 2019 to November 2019 (RDT cohort). Convenience sampling of the first episode per patient was conducted during the specified study period. The EMR used at the institution was Sunrise Clinical Manager (Allscripts, Chicago, IL). Our study was approved by the Rutgers Institutional Review Board and was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Our study did not include factors that required patient consent.

Participants

We included hospitalized adults (≥18 years) with gram-negative bacteremia, as indicated by a positive blood culture for a gram-negative pathogen, who received at least 72 hours of systemic antibiotics. A period of 72 hours was chosen because final culture susceptibilities are usually reported within that period. Patients were excluded if they were transferred from an outside facility because the choice of antibiotics would be based on culture results from the outside facility, resulting in inaccurate timing of antibiotics. Polymicrobial infections, defined as >1 pathogen identified in 1 set of blood cultures, were also excluded because other culture results may influence the provider’s decision to continue or modify antimicrobial therapy. Patients who died or were discharged prior to culture results were also excluded.

Reporting from the microbiology laboratory

The microbiology laboratory is located at Robert Wood Johnson University Hospital. The blood-culture sets consisted of 1 BD BACTEC Plus and 1 BD BACTEC Lytic blood culture bottle (Becton Dickinson, Franklin Lakes, NJ), each ideally filled with 8–10 mL blood. Incubation was performed using the BACTEC FX instrument and Gram-stain results from all single positive blood cultures were called to the nurse caring for the patient, who then informed the patient’s physician. Final culture results with identification of organism and susceptibility to antimicrobials are available in the EMR. This system of reporting occurred during both the historical and intervention periods.

Microbiology procedure

Historical period. Upon identification of a gram-negative organism on the Gram stain of a signal positive blood-culture bottle, specimens were plated on blood, chocolate, and MacConkey agar plates. Approximately 24 hours later, bacterial colonies were identified using the Bruker MALDI-TOF instrument (Bruker, Billerica, MA) and antimicrobial susceptibility testing (AST) was performed using the BD Phoenix automated identification and susceptibility testing system (Becton Dickinson). Final identification and AST results were reported in the EMR ∼48–72 hours after blood culture bottles signaled positive for growth. No antimicrobial stewardship interventions for gram-negative bacteremia were in place during this period.

Intervention period. In May 2019, the microbiology laboratory implemented the Accelerate PhenoTest system (Tucson, AZ), which provides identification and susceptibility results from blood cultures for a limited number of microorganisms within ∼8 hours. The baseline procedures were continued as noted. The results were reported in the EMR. Updated results with additional antibiotic susceptibility results from the Phoenix were entered into the EMR when they became available. If there was a discrepancy between the results of the 2 instruments, additional testing was performed at the discretion of the microbiology laboratory directors and the antimicrobial stewardship response team (ASRT) notified the primary team of the final results. In addition, the microbiology technologists emailed the organisms’ identifications and susceptibility results to the ASRT via electronic communication 24 hours a day and 7 days a week. A phone call was also placed to the ASRT during business hours on weekdays (excluding holidays) and at all other times to the nurse caring for the patient. The ASRT consisted of infectious diseases (ID) faculty members, including physicians and an ID pharmacist, who were responsible for reviewing all gram-negative culture results from the Accelerate PhenoTest and making recommendations to the primary team responsible for the patient. The ASRT also reviewed other RDT results that were not bacteria related and would not affect the patient population studied. If results were reported outside regular business hours, they were e-mailed to the ASRT, who would communicate the results and recommendations the following day. During the RDT period, the microbiology laboratory added personnel to the evening shift, so results were reported during evening hours. Educational efforts were made to all ordering providers prior to the implementation via in-person sessions and electronic distribution through the medical staff newsletter. We implemented RDT and ASRT simultaneously based on experiences at our institution that demonstrated physicians responded to direct prompts to act on results in a timely manner. ^{Reference Bhowmick, Kirn and Hetherington17}

Outcomes

The primary outcome was the total duration of empiric antibiotic therapy for gram-negative bacteremia from the time of the Gram stain. Secondary outcomes included modification of therapy within 24 hours of the reported susceptibilities. Modification was defined as either a de-escalation to a targeted agent or escalation of therapy to broader coverage based on the RDT results. De-escalation was defined as a change in the antibiotic from a broad-spectrum agent to a more targeted agent based on the RDT results. Escalation was defined as a change in the antibiotic to one with a broader spectrum based on the RDT results. Broad-spectrum antibiotics consisted of, but were not limited to, piperacillin-tazobactam, cefepime, meropenem, aztreonam, and ceftazidime, whereas narrow-spectrum antibiotics included, but were not limited to, ceftriaxone, cefazolin, ampicillin, and levofloxacin. Notably, many patients were not started on broad-spectrum antibiotics, and for many patients, targeted therapy did not include a narrow-spectrum antibiotic. All outcomes measuring time were measured from the time of Gram stain (“time zero”). We chose Gram stain rather than blood-culture draw to ensure accurate time measurements for antibiotic changes based on results reported by the microbiology laboratory and to appropriately assess the impact of the RDT. Antibiotics were not assessed prior to Gram stain results. Other outcomes included time to directed therapy, modification within 24 hours of susceptibility results, modification at any point in therapy, modified antibiotic from empiric choice if the antibiotic was modified from empiric therapy, hospital LOS, 14-day in-hospital mortality, subsequent infection with a multidrug-resistant organism (MDRO) within 30 days of the initial positive culture, and development of CDI within 30 days of initial positive culture. Time to directed therapy was defined as the time from the Gram stain result report to the time of escalation or de-escalation to an antibiotic with in vitro activity against the pathogen, and only included patients who had a modification of therapy. Clostridioides difficile testing was performed using a 2-step algorithm using GDH-toxin A/B EIA with reflex to PCR if GDH positive and toxin negative. The process was the same during both the historical and intervention periods. MDRO was defined as a pathogen that had resistance to ≥3 classes of antibiotics.

Statistical analysis

Continuous data are reported as means with standard deviations or medians with interquartile ranges (IQR), as appropriate. All categorical data are reported as percentages. Statistical significance for continuous data was determined using the Student t test for parametric variables or the Wilcoxon rank-sum test for nonparametric variables. Categorical data were analyzed using the χ² or the Fisher exact test, as appropriate. The significance level was determined as a 2-sided P value of <.05. A subgroup analysis of the baseline characteristics was performed for patients who did not achieve the primary outcome. Data were analyzed using R software version 1.4.11.06 software (R Foundation for Statistical Computing, Vienna, Austria).