Timing of appropriate empirical antimicrobial administration and outcome of adults with community-onset bacteremia

Background Early administration of appropriate antimicrobials has been correlated with a better prognosis in patients with bacteremia, but the optimum timing of early antibiotic administration as one of the resuscitation strategies for severe bacterial infections remains unclear. Methods In a retrospective cohort study, adults with community-onset bacteremia at the emergency department (ED) were analyzed. Effects of different cutoffs of time to appropriate antibiotic (TtAa) administration after arrival at the ED on 28-day mortality were examined, after adjustment for independent predictors of mortality identified by multivariate regression analysis. Results Among 2349 patients, the mean (interquartile range) TtAa was 2.0 (<1 to 12) hours. All selected cutoffs of TtAa, ranging from 1 to 96 hours, were significantly associated with 28-day mortality (adjusted odds ratio (AOR), 0.54–0.65, all P < 0.001), after adjustment of the following prognostic factors: fatal comorbidities (McCabe classification), critical illness (Pitt bacteremia score (PBS) ≥4) on arrival at the ED, polymicrobial bacteremia, extended-spectrum beta-lactamase-producer bacteremia, underlying malignancies or liver cirrhosis, and bacteremia caused by pneumonia or urinary tract infections. The adverse impact of TtAa on 28-day mortality was most evident at the cutoff of 48 hours, as the lowest AOR was identified (0.54, P < 0.001). In subgroup analyses, the most evident TtAa cutoff (i.e., the lowest AOR) remained at 48 hours in mildly ill (PBS = 0; AOR 0.47; P = 0.04) and moderately ill (PBS = 1–3; AOR 0.55; P = 0.02) patients, but shifted to 1 hour in critically ill patients (PBS ≥4; AOR 0.56; P < 0.001). Conclusions The time from triage to administration of appropriate antimicrobials is one of the primary determinants of mortality. The optimum timing of appropriate antimicrobial administration is the first 48 hours after non-critically ill patients arrive at the ED. As bacteremia severity increases, effective antimicrobial therapy should be empirically prescribed within 1 hour after critically ill patients arrive at the ED.

between bacteremia onset and antimicrobial administration, ranging widely from <24 hours to <48 hours in previous reports [7,10,11]. To our knowledge, the Surviving Sepsis Campaign (SCC) "International guidelines for the management of severe sepsis and septic shock" recommend that an appropriate antimicrobial therapy be administered within the first 1 hour of recognition of severe sepsis or septic shock [3]. However, focusing on patients with bacteremia, the question "what priority is early enough for appropriate antibiotic administration?" has been debated. Therefore, to describe the beneficial effects of appropriate empirical antimicrobial administration on clinical outcome and to further achieve the "optimum timing" of appropriate antimicrobial administration, we compared the outcome impact of different time cutoffs from emergence department (ED) triage to appropriate antimicrobial administration for adults with community-onset bacteremia.

Study design
This retrospective cohort study was conducted from January 2008 to December 2013 at the ED of a medical center in southern Taiwan. The study hospital is a 1200bed, university-affiliated medical center with an annual ED census of approximate 70,000 patients. The study was reported by the format recommended by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement; partial clinical information on this study cohort has been published [10][11][12][13].
During the study period, the blood cultures sampled at the ED were screened for bacterial growth by a computer database. Clinical information on adults with bacteremia was retrieved from medical charts. In the patients with multiple episodes of bacteremia, only the first episode of each patient was included. Patients were excluded if they had hospital-onset bacteremia, contaminated blood cultures, bacteremia diagnosed prior to visiting the ED, inappropriate dosage or route of antimicrobial administration during all hospitalization, or incomplete clinical information.

Data collection
A predetermined form was used to collect demographic and clinical characteristics, including age, vital signs on arrival at the ED, comorbidities, type of antimicrobial agents administered and duration of administration, bacteremia source, length of hospital stay, bacteremia severity (Pitt bacteremia score (PBS)) on arrival at the ED, comorbidity severity (McCabe classification), and patient outcome. Of all eligible patients, information was retrieved by retrospective review of medical charts, which was done by two authors, and any discrepancy was resolved by discussion between the authors. The primary endpoint was crude mortality within 28 days after onset of bacteremia (arrival at the ED).

Definitions
Community-onset bacteremia indicates that the place of onset of the episode of bacteremia is the community, which includes long-term healthcare facility-acquired and community-acquired bacteremia, as previously described [11,14]. Polymicrobial bacteremia was defined as the isolation of more than one microbial species from a single bacteremia episode. Blood culture samples with potentially contaminating pathogens were considered to be contaminated in accordance with the previous criteria [15].
As previously described [10,11], the antimicrobial therapy was considered appropriate when all the following criteria were fulfilled: (1) the route and dosage of antimicrobial agents were administered as recommended in the Sanford Guide [16] and (2) the bacteremia pathogens were susceptible in vitro to the administrated antimicrobial agent based on the contemporary breakpoints of the Clinical and Laboratory Standards Institute (CLSI) [17]. The time to appropriate antibiotic (TtAa) measured in hours was defined as the period between the arrival at the ED (i.e., ED triage) and the administration of appropriate antimicrobials [18]. The bacteremia severity was graded according to the PBS using a previously validated scoring system based on vital signs, use of inotropic agents, mental status, receipt of mechanical ventilation, and cardiac arrest [11,14]. Comorbidities were defined as described previously [19] and malignancies included hematological malignancies and solid tumors. The prognosis of preexisting diseases was assessed by a previous delineated classification system (McCabe classification) [20]. Bacteremia sources were determined clinically on the basis of the previous definitions [11]. Crude mortality was used to define death from all causes.

Microbiological methods
Blood cultures were incubated in a BACTEC 9240 instrument (Becton Dickinson Diagnostic Systems, Sparks, MD, USA) for 5 days at 35°C. Bacteremia aerobic isolates were prospectively collected during the study period. The bacterial species was identified by means of the Vitek 2 system (bioMérieux, Durham, NC, USA), and antimicrobial susceptibility was determined by the disk diffusion method, based on the performance standards of CLSI in 2016 [17]. The tested drugs included ampicillin/sulbactam, piperacillin/tazobactam, cefazolin, cefuroxime, cefotaxime, ceftazidime, cefepime, ertapenem, imipenem, and levofloxacin. If the patient was empirically treated by other agents, the susceptibility of the indicated agent was measured. For Escherichia coli, Klebsiella species, and Proteus mirabilis (EKP), extended-spectrum beta-lactamase (ESBL) production was detected by the phenotypic confirmatory test with the cephalosporinclavulanate combination disks recommended by the previous CLSI guidelines in 2009 [21].

Statistical analyses
Statistical analyses were performed using the Statistical Package for the Social Sciences for Windows (Version 20.0; Chicago, IL, USA). Clinical data, demographic data, severity, and patient outcomes were compared using the Fisher exact test or the Pearson chi-square test for categorical variables and the independent samples t test or the Mann-Whitney U test for continuous variables. To assess the independent predictors with adjusted odds ratios (AORs), the variables in the univariate analysis with a P value <0.05 were included in a stepwise and backward multivariable logistic regression model. Several TtAa cutoffs were selected: 1 hour, 6 hours, 12 hours, 24 hours, 48 hours, 72 hours, and 96 hours. Cox regression was used to study the impact of varied TtAa cutoffs on clinical outcomes after adjustment for independent risk factors for 28-day mortality.
As all available data were used to maximize the power, no sample size calculation was performed. As suggested, at least 8-10 events per variable are required for reliable multiple logistic regression analysis [22]. For missing data, a complete case analysis was conducted if the percentage of missing values was <5%. If the percentage of missing values was ≥5%, appropriate imputation was performed [23]. A two-tailed P value <0.05 was considered significant.

Baseline characteristics of fatal and surviving patients
Univariate analyses were used to compare clinical variables among adults with fatal and surviving outcome ( Table 1). Nursing-home residents, male gender, critical illness (PBS ≥4) on arrival at the ED, initial syndromes of severe sepsis or septic shock on arrival at the ED, fatal comorbidities (McCabe classification), polymicrobial bacteremia, K. pneumoniae, Pseudomonas species, or ESBL-producing EKP bacteremia, bacteremia pneumonia, underlying neurological diseases or liver cirrhosis, thrombocytopenia, or elevated serum levels of Creactive protein were more frequently noted in patients who died. In contrast, E. coli bacteremia, bacteremia due to urinary tract infections, biliary tract infections, or liver abscess, or underlying hypertension, were more often present in survivors at 28 days after bacteremia onset.

Bacteremia isolates and susceptibility
Because of 240 (10.2%) events of polymicrobial bacteremia and 2109 monomicrobial bacteremia, a total of 2655 causative microorganisms from 2349 patients were collected. The ten leading microorganisms included E.

Predictors of 28-day mortality
The association between 28-day crude mortality and clinical variables, including age, sex, bacteremia severity, bacteremia source, comorbidity severity, causative microorganisms, and major comorbidities was examined by univariate analysis of 2349 adults. The following variables were positively associated with 28-day mortality: having male gender (odds ratio (OR) 1.52; P < 0.001), being resident in a nursing home

Impact of different TtAa cutoffs
Of 2349 patients with community-onset bacteremia, the mean (interquartile range) TtAa was 2.0 (<1 to 12) hours. The majority (1663 patients, 70.8%) were treated by appropriate antimicrobials within 1 hour of arrival at the ED (Table 3). All selected TtAa cutoffs, ranging from 1 hour to 96 hours, had prognostic significance (P < 0.001). In other words, those receiving appropriate antibiotics within the cutoff had a lower mortality rate than those treated by appropriate antibiotics after the cutoff time. The associated AORs ranged from 0.54 to 0.65. Of importance, the lowest AOR (0.54) was evidenced in the TaAa cutoff of 48 hours.

Impact of different TtAa cutoffs on patients with different severity of bacteremia
According to the severity of bacteremia on arrival at the ED, all patients were categorized into three groups: mildly ill (PBS = 0; 633 patients, 26.9%), moderately ill (PBS = 1-3; 1240 patients, 52.8%), and critically ill (PBS ≥4; 476 patients, 20.2%), as shown in Table 4. In mildly ill patients, after adjustment of independent predictors of 28-day mortality, only the TtAa cutoff of 48 hours (AOR = 0.47; P = 0.04) was significantly associated with 28-day mortality (Fig. 1a). Similarly, in moderately ill patients, three TtAa cutoffs (1 hour, 24 hours, and 48 hours) were significantly related to 28-day mortality, and the lowest AOR (0.55) was noted for the TtAa cutoff of 48 hours (Fig. 1b). Focusing on critically ill patients, all TtAa cutoffs were linked to 28-day mortality and the lowest AOR (0.56) was identified for the TtAa cutoff of 1 hour (Fig. 1c).

Discussion
Bacteremia is associated with substantial morbidity and mortality and early interventions are often emphasized [1,14]. However, the estimates of potential benefits of appropriate empirical antibiotic treatment vary widely in the literature, from no effect [8,9] to significant reduction in fatality with an odds ratio of up to 6 [4][5][6][7]. Such a controversy may be related to variations in sepsis severity, comorbidities and the immune status of study cohorts, and the distribution of causative microorganisms [24]. Similar to previous reports dealing with the crucial relationship between patient outcomes and the appropriateness of empirical antimicrobial therapy for specific pathogens, such as Gram-negative bacilli [25], E. coli [26,27], K. pneumoniae [27], S. aureus [28], or bloodstream infections [5,6], our results support that appropriate empirical antimicrobial therapy reduces short-term mortality in patients with community-onset bacteremia. Of importance, in accordance with increasing bacteremia severity, appropriate antimicrobials should be given as soon as possible. Several investigations have attempted to answer the question "How early is enough to administer appropriate antimicrobials in patients with severe infection?" However, different TtAa cutoffs were determined in heterogeneous patient populations reported in the literature. First, for patients with severe Legionella pneumophila pneumonia, delays >8 hours between appropriate antimicrobial administration and ICU AOR adjusted odds ratio, CI confidence interval, ED emergency department, ESBL extended-spectrum beta-lactamase, EKP Escherichia coli, Klebsiella species, and Proteus mirabilis admission resulted in increased mortality [29]. Second, in another study of patients with cancer and septic shock, a delay of >2 hours from ICU admission was associated with a poor prognosis [30]. Finally, in a study of ED patients with severe sepsis and septic shock, a TtAa <1 hour led to a better short-term survival rate [18]. To our knowledge, the present study was the first one focusing on community-onset bacteremia to prioritize antimicrobial therapy in the initial therapeutic strategy. Similar to a previous report of hospital-based bloodstream infection [7], the "optimum timing" of appropriate antimicrobial administration was determined as the first 48 hours after arrival at the ED. Furthermore, consistent with the SSC recommendations [3], this time should be shortened to 1 hour in critically ill patients. However, the administration of in vitro effective antimicrobials to treat bloodstream infections should be as early as possible to achieve optimal treatment of bacterial infections. Any delay in the administration of appropriate drugs should be avoided, if there is a suspicion of severe sepsis or septic shock. This study has several limitations. First, antimicrobial therapy is one of the cornerstones of therapeutic strategies for patients with bloodstream infection. Nevertheless, for critically ill patients with the complications of severe sepsis and septic shock, other elements of early goal-directed therapy, such as central venous pressure, mean arterial pressure, and central venous oxygen saturation, were not evaluated as covariates in this study. Second, in the assessment of the primary endpoint and analysis of the impact of antimicrobial therapy, we excluded patients with incomplete clinical information and those receiving inappropriate antibiotic dosage or route of administration. As a result, only 16 patients were excluded and thus the selection bias was expected to be minimal. Third, as previous investigations dealing with the effects of the appropriateness of empirical antimicrobial administration on patient outcome [4][5][6][7][8][9], no detailed information was available in our study on early resuscitation. However, the majority (1663, 70.8%) of the 2349 patients with community-onset bacteremia received the first dose of appropriate antimicrobials within one hour of arrival at the ED, and the study population The significance of the effects of different cutoffs of the time to appropriate antibiotic (TtAa) on 28-day mortality were examined by univariate and Cox regression analyses. OR odds ratio, CI confidence interval, AOR adjusted odds ratio. a Pitt bacteremia score ≥4 on arrival at the emergency department. b Adjusted for independent predictors of 28-day mortality recognized in the multivariate regression: critical illness; a fatal comorbidity (McCabe classification); polymicrobial bacteremia; Extendedspectrum beta-lactamase-producing bacteremia; bacteremia because of pneumonia or urinary tract infections; and underlying malignancy or liver cirrhosis were believed to be managed according to the SSC recommendations. Thus, the clinical outcomes of the patients receiving appropriate antimicrobial therapy at different time points after arrival at the ED could be compared with each other. Fourth, as noted before [4][5][6][7][8][9], no analyses of source control were performed in the present study. However, taking the cases of urosepsis as the example, in those with or without interventions for source control, such as percutaneous nephrolithotomy or nephrectomy, the 28-day crude mortality rate was not different (4/48, 8.3% vs. 35/700, 5.0%; P = 0.31), indicative of standard quality of clinical care in the study hospital. Finally, as this study was conducted at one hospital, the critical TtAa cutoffs we identified may not be generalizable to other communities with variable infection sources or severity of community-onset bacteremia. However, our findings highlight that more severely ill patients with bacteremia require earlier administration of appropriate antimicrobials to achieve a favorable outcome.

Conclusions
Our results demonstrate that the elapsed time from arrival at the ED to the administration of appropriate antimicrobials is a crucial determinant of short-term outcomes in patients with community-onset bacteremia. The optimum timing of appropriate antimicrobial administration in mildly ill patients was the first 48 hours after arrival at the ED. With increasing severity of bacteremia, this time should be limited to the first 1 hour in critically ill patients. Therefore, to achieve a favorable outcome in critically ill patients, epidemiological surveillance, rapid pathogen identification, or the incorporation of broadspectrum antimicrobials as empirical therapy into an antibiotic stewardship program should be considered.  Fig. 1 Adjusted odds ratios of different cutoffs of the time-toappropriate antibiotic for 28-day crude mortality in adults, categorized by the Pitt bacteremia score on arrival at the emergency department in mildly ill (a), moderately ill (b), and critically ill (c) patients, using Cox regression after adjustment for the independent predictors of 28-day mortality (including a fatal comorbidity (McCabe classification), polymicrobial bacteremia, extended-spectrum beta-lactamaseproducing bacteremia, bacteremia because of pneumonia or urinary tract infections, and underlying malignancies or liver cirrhosis). *P value <0.05; **P value <0.001