Discussion

Four risk factors were identified in our study that increased the odds of encountering primary resistance to ceftriaxone (EC or KP) or cefepime (ENC or PS). These factors included mechanical ventilation, admission from SNF/CCH or infection within the last 6 months, history of connective tissue disease, and history of hemiplegia. Initially, a score was developed to characterize the risk of resistance; however, our goal was to identify the subpopulation with the lowest risk of resistance to justify the use of first-line antimicrobial agents.

A previous study was performed in 192 adults with cancer who developed Enterobacter bacteremia. The identified species included E. cloacae (67.7%), Klebsiella (formally Enterobacter) aerogenes (29.2%), and four other Enterobacter sp. Of the Enterobacter isolates, 27.6% were reported to be resistant to extended-spectrum cephalosporins (ESC) and the remainder were susceptible. The authors found that previous exposure to an ESC, tumor progression, recent surgery, and nosocomial acquisition were risk factors for ESC resistance. Lower respiratory tract infection, tumor progression, septic shock, and E. aerogenes bacteremia were identified as risk factors for mortality. ^{Reference Huh, Kang and Kim13} In contrast, we assume that E. cloacae and K. aerogenes are resistant and report as such, and cefepime is an acceptable therapy if the isolate is susceptible in vitro. Another study included patients with community-onset Escherichia coli bacteremia in Taiwan. The incidence of ESC resistance was 9.6% in bacteremia isolates. Among the ESC-resistant isolates, appropriate empiric therapy was administered to 45.9% of 133 patients, compared to 92.6% of 543 patients with susceptible isolates. ESC resistance was associated with increased 14-day mortality and longer length of stay. Factors associated with ESC resistance included hospitalization within the past year, exposure to antibiotics within the past 15 days, residence in a long-term care facility, underlying genitourinary disease, and presence of an implantable intravenous port. ^{Reference Lin, Huang, Wang, Chen and Chang14}

Antimicrobial stewardship requires a balance between selecting active empiric therapy and avoiding overuse of protected antimicrobial agents. Ensuring active empiric therapy is particularly important for patients with septic shock. ^{Reference Sloan, Rodriguez and Seward15} Many antibiotics, including ceftriaxone and cefepime, qualify as broad-spectrum antibiotics and are appropriate for empirical coverage. Among a large population of surgical patients, patients who received an appropriate initial antibiotic treatment (based on susceptibility to the pathogen(s)) did not demonstrate improved survival over patients who had a pathogen resistant to the empiric therapy. ^{Reference Davies, Efird and Guidry16} In 670 patients with Gram-negative bacteremia, inappropriate empiric treatment (pathogen resistant) was not significantly associated with 7-day or 30-day all-cause, mortality. ^{Reference Fitzpatrick, Biswas and Edgeworth17} Likewise, patients with community-onset bacteremia due to ESBL-producing E. coli or K. pneumonia did not exhibit increased mortality as long as they received a carbapenem as definitive therapy. ^{Reference Joo, Park and Lee18} In sepsis due to a urine source caused by E. coli susceptible or resistant to a third-generation cephalosporin, mortality was higher in the group with resistant isolates; however, mortality was not associated with inappropriate empiric therapy. ^{Reference Tocut, Zohar, Schavartz, Yossepowitch and Maor19} In a systematic review and meta-analysis, inappropriate empirical antibiotic therapy was associated with increased mortality, but 42.3% of patients had septic shock. The odds ratio for mortality was 22.3 for presence of septic shock and 1.83 for inappropriate empiric therapy. ^{Reference Rojas, Palacios-Baena, Lopes-Cortes and Rates20} Additionally, another study demonstrated that inappropriate empiric treatment contributes to worse outcomes in patients with hospital acquired pneumonia, ventilated hospital acquired pneumonia, and ventilator-associated pneumonia. The organism with the highest incidence of inappropriate empiric therapy was P. aeruginosa (24.2%). ^{Reference Zilberberg, Nathanson, Puzniak, Dillon and Shorr21}

The goal of this study was to identify a subpopulation at low risk for primary resistance, given the more common Gram-negative organisms recovered from blood cultures. Patients with urinary tract infections and intra-abdominal infections are most likely to have Enterobacterales infections (eg, E. coli and K. pneumoniae), and ceftriaxone could be selected in the absence of risk factors (Table 4). If the patient had risk factors for pseudomonal infection or hospital-acquired Gram-negative infection, cefepime could be selected in the absence of risk factors. Owing to the high early mortality risk, we suggest not applying this recommendation to patients with septic shock. The restriction of carbapenem use has already been implemented in many institutions. Carbapenems are often restricted to use in patients with worsening infection when receiving a cephalosporin or piperacillin-tazobactam, documented resistance to empiric therapy, history of ESBL-producing Gram-negative infection, and/or approval by an infectious disease specialist or antimicrobial stewardship program. ^{Reference Goodlet and Nailor22} Another strategy is to require infectious disease consultation for extending carbapenem use beyond 72 hours. Interestingly, this strategy resulted in a 45% to 65% reduction in the initial use of a carbapenem. ^{Reference Mani, Lan and Jain23}

As this study was based on a retrospective review of the electronic health records, a noteworthy limitation is that of missing or inaccurate data. Specifically, prior hospitalizations may have taken place in a different hospital system, or there may be missing past medical history. This study focused on the four most common pathogens of Gram-negative bacteremia but did not capture data on patients with other Gram-negative organisms. We fell short of the planned number of cefepime-resistant ENC isolates. We hypothesized that EC, KP, and ENC would be representative of other Enterobacterales organisms. PS presents a common nonfermenter with resistance usually caused by a mixture of reduced permeability, ampC beta-lactamase hyperproduction, and efflux. The model will need to be confirmed in a prospective study or at least using a different population. However, given that the majority of patients do not have any of the identified risk factors, many can be managed with first-line antimicrobial agents.

Ultimately, our results have implications for limiting carbapenem use. There is a small population of patients admitted to the intensive care unit without identified risk factors for resistance, but with a high risk of mortality within the first 24–48 hours. This group of patients may need expanded coverage initially until susceptibility to first-line agents can be confirmed. Most of our cohort did not have a history of infection with a resistant organism, but a history of infection within the previous 6 months was identified as a risk factor for primary resistance. Admission from a chronic care hospital or SNF was associated with similar risk. For patients without a high risk of early mortality (eg, absence of septic shock), we suggest that empiric therapy with first-line antimicrobial agents be strongly considered. Organisms can be identified in many cases shortly after growth in culture, although susceptibility testing can take an additional 24–48 hours. At this point, the patient has received empiric antimicrobial therapy for about a day, and there can be clinical decisions regarding the response to treatment and susceptibility. Our study can also inform the design and implementation of clinical decision support tools to limit the overuse of broad antimicrobials, such as carbapenems.