Evaluating vancomycin and piperacillin-tazobactam in ED patients with severe sepsis and septic shock

a b s t r a c t

Study objective: To determine the frequency and cause of inadequate initial antibiotic therapy with vancomycin and piperacillin-tazobactam in patients with severe sepsis and septic shock in the emergency department (ED), characterize its impact on patient outcomes, and identify patients who would benefit from an alternative initial empiric regimen.

Methods: Retrospective cohort study conducted between 2012 and 2015 in which 342 patients with culture- positive severe sepsis or septic shock who received initial vancomycin and piperacillin-tazobactam were reviewed to determine appropriateness of antimicrobial therapy, risk factors for Inappropriate use, and outcome data. Univariate and multivariate regression analyses were determined to identify associations between inappro- priate antibiotic use and outcomes and to identify risk factors that may predict which patients would benefit from an alternative initial regimen.

Results: Vancomycin and piperacillin-tazobactam were inappropriate for 24% of patients with severe sepsis or septic shock, largely due to non-susceptible infections, particularly ESBL organisms and Clostridium Difficile. Risk factors included multiple sources of infection (OR 4.383), admission from a skilled nursing facility (OR 3.763), a history of chronic obstructive pulmonary disease (COPD) (OR 3.175), intra-Abdominal infection (OR 2.890), and immunosuppression (OR 1.930). We did not find a mortality impact.

Conclusion: Vancomycin and piperacillin-tazobactam were an inappropriate antibiotic combination for approxi- mately 24% of patients with either severe sepsis or septic shock in the ED. Patients with known COPD, residence at a skilled nursing facility, a history concerning for Clostridium difficile, and immunosuppression would benefit from an alternative regimen. Future prospective studies are needed to validate these findings.

Introduction

Timely administration of appropriate, Intravenous antibiotic therapy in septic patients has been associated with significant reductions in mortality, hospital length of stay, and organ dysfunction in multiple studies [1-6]. However, what defines “appropriate” antimicrobials is not as clear in the emergency department (ED) when culture data is not available and the source of infection is not always known. The most recent update to the Surviving Sepsis Campaign guidelines reiterates the importance of antimicrobial therapy broad enough to cover all likely pathogens, with consideration for site of infection, un- derlying comorbidities, history of immunosuppression, recent infection or colonization, and antimicrobial use in previous three months, among

* Corresponding author at: Department of Pharmacy, University of California, San Diego Health, 200 West Arbor Drive, M-8765, San Diego, CA 92103, United States.

E-mail address: [email protected] (C. Le).

other risk factors (strong recommendation, moderate quality of evi- dence) [7]. In our practice, a sepsis protocol guides antimicrobial selec- tion in the ED based on suspected source of infection, local susceptibility patterns, and the setting in which the infection is believed to have orig- inated. Over time, combination Empiric therapy with vancomycin and piperacillin-tazobactam has become a standard regimen at our institu- tion due to broad antimicrobial coverage and familiarity of use. Howev- er, it is not possible to determine the causative organisms of sepsis while patients are in the ED, and this combination of vancomycin and piperacillin-tazobactam may not be ideal in all clinical scenarios.

Evidence is lacking to identify what patient populations may not benefit from the combination of a standardized antibiotic regimen with suspected severe sepsis or septic shock in the ED. Investigation of this topic is especially important to improve empiric antimicrobial se- lection in sepsis in light of evidence that showed up to a fivefold in- crease in mortality among septic shock patients initiated on inadequate antimicrobial regimens [8]. Our primary objective is to

https://doi.org/10.1016/j.ajem.2017.12.055

quantify the frequency in which the administration of intravenous vanco- mycin and piperacillin-tazobactam Combination therapy for septic pa- tients in the ED was considered inadequate based on pre-defined criteria. Secondary objectives include the impact of appropriateness in patient-centered outcomes, such as in-hospital mortality, length of stay, and duration of Vasoactive medications. We also aim to assess risk factors or clinical characteristics that may be associated with inadequate use.

Methods

Study design and setting

This was a retrospective cohort study of persons aged 18 years and older that presented to one of the two EDs at our institution with either severe sepsis or septic shock. One hospital functions as a safety-net hos- pital in an urban setting, whereas the other is a quaternary referral cen- ter. The combined yearly census between these two EDs is approximately 60,000 patients. Between July 2012 and July 2015, all pa- tients with suspected severe sepsis or septic shock in the ED were reviewed by a senior critical care attending physician and entered into a prospectively enrolled registry. The time period corresponds to the time this database was maintained. Patients were considered to have severe sepsis when two or more systemic inflammatory response syn- drome (SIRS) criteria (heart rate N 90 beats per minute, respiratory rate N 20 breaths per minute, temperature N 38.3 ?C or b 36 ?C, white blood cell count N 12,000 cells/mm³, b 4000 cells cells/mm³, or N 10% im- mature bands) were present, and had either confirmed or suspected in- fection. We used the 2012 Surviving Sepsis Campaign guidelines for the definitions of severe sepsis and septic shock [9]. Individual patient charts were then retrospectively reviewed to collect additional hospital admission data, medication administration information, and laboratory results. Ethical approval was obtained from the institution review board at the University of California, San Diego and the Human Research Pro- tections Program with waiver of informed consent.

Selection of participants and definitions

Patients were included if they were 18 years of age or older, met criteria for either severe sepsis or septic shock in the ED, and received only intravenous vancomycin and piperacillin-tazobactam as initial an- timicrobial therapy while in the ED. This was defined a priori as at least one dose of intravenous vancomycin and one dose of piperacillin- tazobactam as the first two antimicrobial agents. Patients who received oral vancomycin were excluded. Patients were still eligible for inclusion if additional antimicrobial agents were administered 2 h after vancomy- cin or piperacillin-tazobactam was given. We chose 2 h as the cut-off for initial vancomycin-tazobactam therapy because it allowed providers adequate time to review the medical record and make any changes to

susceptible infection based upon culture data, history of non- susceptible infection within 6 months, insufficient coverage or non- optimal indication, or allergy. These criteria were selected based upon review of prior published manuscripts and also consensus upon the re- search group as an expansive definition of inadequate therapy covering all potential reasons for inadequate antibiotic therapy [1-3]. Table 1 in- cludes a description of the different classifications of inadequate thera- py as well as examples. Classification of antibiotic resistance was made by guidelines and breakpoints defined by the Clinical Laboratory and Standards Institute during the study period, followed by our institution’s microbiology lab [12]. In the case of a patient with both a non-susceptible infection and a history of a non-susceptible infection, we placed the patient in the non-susceptible infection category. All other cases were considered appropriate. Of note, we were unable to find guidelines or strong evidence defining the time required to clear a resistant organism, thus this selection of 6 months was chosen in con- sensus among members of the research team.

Immunosuppression was defined as recent or chronic corticosteroid use, history of HIV infection with CD4 count b 200 cells/uL, any history of organ transplantation, or current therapy with immunosuppressive medications. The diagnosis of heart failure, COPD, and other comorbid- ities were made based upon review of the subject’s medical record. A patient was considered to have Clostridium difficile if this was felt to be the inciting infection requiring hospitalization, not a secondary infec- tion acquired during the hospital stay.

Methods and measurements

All data were reviewed and retrospectively extracted into a stan- dardized collection sheet devised by the research group and data ab- straction was performed in accordance with previously published standards for chart reviews [13]. Patient cases were then reviewed and classified as “appropriate” or “inadequate” uses of vancomycin and piperacillin-tazobactam. The research group, comprised of emer- gency medicine and critical care physicians and pharmacists, proposed and unanimously agreed upon these conditions to identify those cases in which vancomycin and piperacillin-tazobactam are insufficient or in- adequate as Empiric treatment.

Outcome measures

The primary outcome was the inadequate use of empiric vancomy- cin and piperacillin-tazobactam in patients with severe sepsis or septic

Table 1

Study criteria for inadequate use of vancomycin and piperacillin-tazobactam.

Condition Definition Example

the initial antibiotic regimen. Additional antibiotic therapy after this time point was considered to be a revision to initial antibiotics. There were no identified pre-defined time points we were able to locate in the literature. We excluded patients transferred from another hospital, patients transferred to another acute care, and those on hospice care at the time of admission. Furthermore, patients whose culture data returned negative for bacterial growth were not included, based on methodology in previously published manuscripts on similar topics [10,11].

“Appropriate” antibiotic use was defined a priori as in vitro suscepti- bility on culture data obtained during the present admission to either vancomycin or piperacillin-tazobactam. We included only patients with culture data that was felt to be the causative organism responsible for severe sepsis or septic shock by the treating team and reviewed data

Non-susceptible infection

History of

non-susceptible infection

Non-optimal indication

Culture returns positive for bacteria neither vancomycin nor piperacillin-tazobactam show in vitro activity against

Culture from previous six months grew bacteria non-susceptible to vancomycin or piperacillin-tazobactam

Current clinical guidelines do not recommend vancomycin and piperacillin-tazobactam as preferred empiric treatment for type of infection

Blood culture drawn during current admission is positive for extended-spectrum

beta-lactamase producing Klebsiella pneumoniae Sputum culture taken during previous admission from three months ago grew Enterococcus faecium that showed resistance to vancomycin

Positive lumbar puncture results confirm central nervous system (CNS) infection in patient clinically suspected to have CNS infection

from blood cultures, sputum cultures, urine cultures, wound and opera-

tive cultures, cerebrospinal fluid cultures, and Clostridium difficile poly- merase chain reaction (PCR) data. Our study defined “inadequate” use as the presence of at least one of the following conditions: non-

Allergy Documented allergic reaction to vancomycin,

piperacillin-tazobactam, or component of either product

Patient with no documented allergy history develops hives after administration of piperacillin-tazobactam

shock based on the above-defined criteria. Secondary outcomes includ- ed reason for inadequate use, characterization of non-susceptible infec- tions, in-hospital mortality, hospital length of stay, admission to intensive care unit (ICU), ICU length of stay, initiation of vasopressor therapy, and length of vasopressor therapy.

Primary data analysis

Descriptive statistics were used to describe baseline characteristics and study outcomes. Independent t-tests were used to compare contin- uous variables, while Chi-squared and Fisher’s exact tests were used for categorical variables. A two-sided alpha of b 0.05 was considered statis- tically significant. We conducted a backwards step-wise multivariable regression analysis to further characterize possible predictors of inade- quate antibiotic use and control for potential confounders. An entry probability of F set at 0.10 for removal was used. We began with a model containing all variables that had shown statistical significance dur- ing univariate analysis (p b 0.05). We then excluded variables one at a time starting with the variable with the highest p-value in the model, until only statistically significant variables remained in the model. We then selected specific variables suspected to have biological relationships with the outcome of interest and reinserted them into the model. None were found to have statistically significant explanatory value. Due to the high prevalence of Clostridium difficile infection in patients with intra-abdominal infections, we repeated our model excluding the pa- tients with C. difficile infection with the same approach described above. All analyses were performed with SPSS (version 24.0; IBM, Armonk, NY) and STATA (version 14; StataCorp, College Station, TX).

Results

Characteristics of study subjects

A total of 800 patients met criteria for severe sepsis or septic shock during the study period. Of these, 545 (68%) received vancomycin and

piperacillin-tazobactam as their initial antimicrobial and were evaluat- ed for inclusion. From this group, 203 patients were excluded: 2 patients were transferred from an outside acute care facility, 4 were transferred to an outside acute care facility for continuation of care, and 197 pa- tients had cultures return negative for bacterial growth (Fig. 1). As a re- sult, 342 patients were included.

Baseline characteristics of the included patients are displayed in Table 2. Most patients (n = 243; 71%) were admitted from a private res- idence and presented with severe sepsis (n = 209; 61%). The most com- mon underlying comorbidities were diabetes, history of malignancy, chronic kidney disease, and heart failure, while infections primarily in- volved the urinary, respiratory, or gastrointestinal systems. Patients ad- mitted from a skilled nursing facility were more likely to have received inadequate initial use of vancomycin (37.3% vs. 16.6%, p = 0.000). Fur- thermore, the prevalence of heart failure, chronic obstructive pulmo- nary disease (COPD), and gastrointestinal or intra-abdominal source of infection was much higher in the inadequate use group and all differ- ences were statistically significant. Importantly, there was no difference in the time to antibiotic administration, initial amount of fluid resuscita- tion administered or initial lactate levels between groups.

Main results

Review of the 342 patient cases showed most patients received van- comycin and piperacillin-tazobactam appropriately (n = 259; 76%) whereas 24% (n = 83) of patients received inadequate initial antibiotic therapy with the principal reason being infection with a non- susceptible organism (n = 72; 87%), followed by a history of an infec- tion with a non-susceptible organism in the past six months (n = 8; 9.6%). Rarely were vancomycin and piperacillin-tazobactam used for non-optimal indications or in patients with a history of allergy to the ad- ministered antibiotic (Table 3). Further investigation revealed that, within the non-susceptible infection subgroup (n = 72), 33% of these patients (n = 24) had culture data from the past six months growing non-susceptible organisms (Fig. 1). These positive culture results were

Study population

N = 352

Appropriate initially and in retrospect

Nonsusceptible infection

Non-optimal indication

N = 2

Allergy

N = 1

Previous cultures show nonsusceptible organism

N = 24

No previous cultures show nonsusceptible organism

N = 48

Inadequate initially and in retrospect

Appropriate initially, but inadequate in retrospect

Fig. 1. Characterization of inadequate uses by reason.

Table 2

Demographic and baseline characteristics of study patients.

Characteristic Appropriate

use

(N = 259)

Mean age +- SD - yr	63 +- 16	63 +- 17	0.957
Male sex - no. (%)	162 (62.5)	42 (50.6)	0.071
Admission source - no. (%)			0.000
Private residence	192 (74.1)	51 (61.4)	0.036
Skilled nursing facility	43 (16.6)	31 (37.3)	0.000
Homeless	19 (7.3)	1 (1.2)	0.055
Incarcerated	0	5 (1.9)	0.341
Comorbidities - no. (%) Diabetes	92 (35.5)	32 (38.6)	0.694
Malignancy history	92 (35.5)	27 (32.5)	0.692
Chronic kidney disease^a	70 (27.0)	26 (31.3)	0.484
Heart failure	45 (17.4)	30 (36.1)	0.001
Immunosuppression	50 (19.3)	23 (27.7)	0.123
Cirrhosis	35 (13.5)	11 (13.3)	1.000
Chronic obstructive pulmonary disease	14 (5.4)	16 (19.3)	0.000
Septic shock - no. (%)	101 (39.0)	32 (38.6)	1.000
Suspected source of infection - no. (%)			0.017

Inadequate use

(N = 83)

p value

“appropriate initially, but inadequate in retrospect”-that is, vancomy- cin and piperacillin-tazobactam were appropriate at time of presenta- tion and only became inadequate when later cultures showed non- susceptibility.

Secondary outcomes

Factors associated with initial vancomycin and piperacillin- tazobactam use that was considered inadequate included admission from a skilled nursing facility, history of COPD or heart failure, and intra-abdominal infection site (Table 5). Notably, 61% (14/23) of the pa- tients who received inadequate initial vancomycin and piperacillin- tazobactam had Clostridium difficile as the causative organism for their sepsis. In the final backward selection multivariate regression analysis, we found that admission from SNF, history of COPD, intra-abdominal or multiple infection, and history of immunosuppression were indepen- dent risk factors for the administration of an inadequate antibiotic reg- imen (Table 5a). In the model that excluded patients with Clostridium difficile, the following patient populations had significant odds for inad-

Urinary tract or genitourinary	86 (33.2)	28 (33.7)	1.000	equate initial antibiotic therapy: admission from a SNF (OR 3.658 (95%
Respiratory	52 (20.1)	15 (18.1)	0.753	CI 1.922, 6.960), p = 0.000), immunocompromise (OR 2.470 (95% CI
Gastrointestinal or intra-abdominal Soft tissue or skin Cardiac	42 (16.2) 26 (10.0) 6 (2.3)	23 (27.7) 4 (4.8) 1 (1.2)	0.025 0.183 1.000	1.258, 4.850), p = 0.009), and history of COPD (OR 2.961 (95% CI 1.308, 6.7703), p = 0.009) (Table 5b).
Central nervous system	2 (0.8)	1 (1.2)	0.567	The overall in-hospital mortality rate for the study population was
Multiple	18 (6.9)	11 (13.3)	0.110	15.2% (n = 52), with comparable rates of 14.3% and 18.1% in the appro-

priate and inadequate use groups, respectively (Table 4). No statistically significant differences existed between the groups in regards to in- hospital mortality, ICU admission, length of hospital or ICU stay, initia- tion of vasopressor therapy, or length of vasopressor therapy.

Other^b	16 (6.2)	0	0.015
Unknown	11 (4.2)	0	0.072
Mean time to first antimicrobial +- SD - min	99 +- 79	91 +- 57	0.413
Mean time to piperacillin-tazobactam +-	103 +- 83	97 +- 62	0.531
SD - min Mean time to vancomycin +- SD - min	127 +- 91	126 +- 77	0.889
Antibiotics given before cultures drawn -	6 (2.3)	2 (2.4)	1.000
no. (%) Mean lactate level +- SD - mmol/L	1.76 +- 1.18	1.82 +- 1.27	0.664
Fluid administration in first 6 h +- SD -	38 +- 19	36 +- 18	0.496

Discussion

mL/kg

^a Including patients with end-stage renal disease with or without intermittent hemodialysis.

^b Other infections included cases of central-line associated bloodstream infection, neutropenic fever, and osteomyelitis, among others.

available to providers at the time antimicrobials were ordered and ver- ified. This subgroup of patients (n = 24) received vancomycin and piperacillin-tazobactam for infection with a non-susceptible organism despite historical cultures showing a non-susceptible organism, and were thus classified as “inadequate initially and in retrospect.” These pa- tients are differentiated from those with only a history of non- susceptible infection and for whom the empiric regimen did show in vitro activity against the current infection (n = 8; 9.6%). The remain- ing 48 patients in the non-susceptible infection group were

Table 3

Instances of inadequate use of initial therapy with vancomycin and piperacillin- tazobactam.

Inadequate criterion met Inadequate use (N

= 83)

Non-susceptible infection - no. (%) 72 (86.7)

We found that 76% of severe sepsis and septic shock cases were treated appropriately with empiric Initial administration of vancomycin and piperacillin-tazobactam. These results are comparable to rates re- ported in previous studies examining appropriateness of antibiotic ther- apy in severe sepsis and septic shock patients in the ICU [10,11]. However, ours is unique in that no prior investigations have focused upon the administration of this commonly used antibiotic regimen to septic patients in the ED. This finding supports that the frequent use of vancomycin and piperacillin-tazobactam empiric therapy in severe sep- sis and septic shock, though imperfect, is still acceptable as initial anti- microbial therapy for the majority of patients in the ED. The rate of administration that was considered inadequate 24% in our total popula- tion, however, was higher than expected and underlies a substantial co- hort of undertreated patients. Reasons for this discrepancy may be multifactorial. Emphasis is now placed on reducing the time to antibiot- ic administration and optimizing fluid resuscitation as part of Sepsis Bundles, especially in the current era of the Joint Commission’s sepsis core measures, rather than appropriate antibiotic selection [14]. Addi- tionally, our institution cares for a significant number of patients under- going transplants, which may increase their exposure to this particular antibiotic regimen.

Inadequate therapy in our study was primarily associated with non- susceptible infection due to bacteria resistant to both vancomycin and piperacillin-tazobactam. Over half the cases were due to infection

with extended-spectrum beta-lactamase (ESBL) producing organisms

Extended-spectrum beta-lactamase (ESBL) producing bacteria

28 (33.7)

and non ESBL-organisms resistant to piperacillin-tazobactam (n = 44;

Non-ESBL organism resistant to piperacillin-tazobactam 16 (19.3)

Clostridium difficile 14 (16.9)

Vancomycin-resistant Enterococcus 6 (7.2)

Resistant to vancomycin 2 (2.4)

Klebsiella pneumoniae carbapenemase-producing bacteria 1 (1.2)

Mycobacterium spp. 1 (1.2)

History of non-susceptible infection - no. (%) 8 (9.6)

Non-optimal indication - no. (%) 2 (2.4)

Allergy - no. (%) 1 (1.2)

53%). This may indicate that while Gram-positive organisms including

MRSA are adequately covered by vancomycin, the spectrum of gram- negative and anaerobic pathogens in our population was not sufficiently treated with piperacillin-tazobactam. In retrospect, broader spectrum agents such as carbapenems or aminoglycosides might have been pre- ferred in this subset of patients. Furthermore, we identified that patients with severe sepsis or septic shock from Clostridium difficile infection made up a significant percentage of cases with inadequate initial

Table 4

Secondary outcomes.

Endpoint	Appropriate use (N = 260)	Inadequate use (N = 83)	p value
In-hospital mortality - no. (%)	37 (14.3)	15 (18.1)	0.386
Admission to ICU - no. (%)	176 (68.0)	66 (79.5)	0.052
Mean length of hospital stay +- SD - day	12.8 +- 15.1	11.1 +- 7.6	0.340
Mean length of ICU stay +- SD - day	4.5 +- 7.7	4.3 +- 5.1	0.822
Vasopressor therapy initiated - no. (%)	114 (44.0)	38 (45.8)	0.801
Mean length of vasopressor therapy +- SD - day	1.8 +- 3.5	1.1 +- 1.7	0.090

therapy. With a rising incidence of Clostridium difficile infections identi- fied in EDs across the United States, providers should note risk factors and findings concerning for this condition when choosing an initial an- tibiotic regimen [15].

Furthermore, 33% of cases in which therapy was considered inade- quate defined by a non-susceptible infection, previous cultures positive for a resistant organism were available to physicians and pharmacists within the electronic medical record system at the time of antimicrobial selection. These positive cultures from the past six months did not nec- essarily grow the same pathogen as that found during this admission, but they were indicative of a history of infection, or at least colonization, with resistant bacteria. Though we do not advocate consistent use of broader-spectrum antibiotics, these findings suggest that reviewing previous culture data-and ordering broader agents accordingly-may be the most impactful and straightforward intervention to reduce inad- equate use of vancomycin and piperacillin-tazobactam.

This finding led us to stratify those patients who received vancomy- cin and piperacillin-tazobactam appropriately at time of presentation (n

= 259 combined with n = 48). Because 48 cases were considered inad- equate in retrospect, the “success” rate of vancomycin and piperacillin- tazobactam when given appropriately at presentation to the ED was ef- fectively 84%: 259 patients received the antimicrobials appropriately both initially at presentation and retrospectively after reviewing final culture data, out of a total 307 initially appropriate uses.

Admittedly, the majority of patients who received vancomycin and piperacillin-tazobactam inadequately could not be predicted: 48 out of 83 patients, or 58%, developed a non-susceptible infection in the ab- sence of historical resistant infections (Fig. 1). In these cases that lack clinical indicators to suggest vancomycin and piperacillin-tazobactam are inadequate, it is imperative that culture results are closely moni- tored and antimicrobials adjusted accordingly.

Examining the differences between the two groups at baseline does, however, reveal significant disparities, namely higher prevalence of heart failure, COPD, GI or intra-abdominal infection, and admission from SNF in the inadequate use group. In fact, when we stratified pa- tients by admission source, 42% of SNF patients receiving the combina- tion of vancomycin and piperacillin-tazobactam were considered inadequate, in comparison to 21% of private resident patients, and 24%

Table 5 Results of multivariate regression analysis to determine independent risk factors associat- ed with inadequate vancomycin and piperacillin-tazobactam use (a) including patients with Clostridium difficile infection, (b) excluding patients with Clostridium difficile

infection.

of all patients in the study. Patients frequently re-hospitalized for heart failure exacerbations or patients transitioning between sub- acute or long-term Acute care facilities may have increased healthcare exposure, leading to the development of resistant infection. As for the high prevalence of GI and intra-abdominal sources of infection in the in- adequate use group, we attributed this trend to those cases of Clostridi- um difficile infection, which were not initially recognized by providers. These findings highlight that underlying comorbidities of heart failure or COPD, presentation from a SNF, and suspected GI or intra- abdominal source of infection, may be independent risk factors that in- crease the likelihood of receiving vancomycin and piperacillin- tazobactam that might be considered inadequate to cover the offending organism. Interestingly, in our multivariate regression analyses (one with all patients and one excluding those with C. difficile) to identify in- dependent risk factors for inadequate use of vancomycin and piperacillin-tazobactam, we found that 3 factors were present in both analyses: admission from a skilled nursing facility, a history of COPD, and immunosuppression. We suspect that this is largely due to the high number of cases of Clostridium difficile present in our patient pop- ulation who initially received initial inadequate vancomycin and piperacillin-tazobactam.

In our population, we encountered very few cases of therapy consid-

ered inadequate due to allergy or non-optimal indication. This likely re- flects the versatility of the vancomycin and piperacillin-tazobactam combination for multiple types of infection, physician accuracy in iden- tifying the infectious source, and consistent practices in obtaining, veri- fying, and reviewing patient allergy history before medication administration.

In evaluating the study’s secondary outcomes, we found no statisti- cally significant differences in in-hospital mortality, admission to ICU, hospital length of stay, ICU length of stay, initiation of vasopressor ther- apy, or days of vasopressor therapy between the appropriate and inad- equate groups. Overall in-hospital mortality in our study population was 15% (n = 52), which, although lower than the 25 to 30% rates cited in other recent studies [16-18], is not wholly representative of all septic patients, as we excluded patients with culture-negative sepsis or non-bacterial infection and patients who received an alternate com- bination of Empiric antibiotics. Though in-hospital mortality was quan- titatively lower in the appropriate use group, this was not a statistically significant difference. A larger population size may show reduction in- hospital mortality or shortened length of hospital and ICU stay, in line with literature reporting improved outcomes with appropriate antibi- otics [8,19-21]. It is important to note that Kumar et al. conducted

their landmark study in the era preceding aggressive, early manage-

Risk factor	Odds ratio	p value	95% confidence interval
a Admission from SNF^a	3.783	0.000	2.037-7.026
History of COPD^b	3.175	0.005	1.420-7.099
Intra-abdominal infection	2.890	0.001	1.514-5.515
Multiple infections	4.383	0.001	1.892-10.152
Immunosuppression	1.930	0.050	1.000-3.725
b Admission from SNF^a	3.66	0.000	1.922-6.960
Immunosuppression	2.47	0.009	1.258-4.849
History of COPD^b	2.961	0.009	1.308-6.703
^a SNF = skilled nursing facility.

ment for sepsis and patients with culture-negative sepsis were includ- ed, which may account for the difference in mortality rates between our studies [8]. Furthermore, our definition of inadequate therapy in- cluded patients with recent culture data showing resistance to vanco- mycin or piperacillin-tazobactam even if the identified organism was susceptible to these agents, which may have artificially increased the survival rate in the inadequate antibiotic use group.

Limitations

As this was a single-center study with a small sample size, our local

^b COPD = chronic obstructive pulmonary disease.

resistance patterns, data, and results may not be representative in other

settings or institutions with different patient populations and antibiograms. Because this was a retrospective study, all results should be considered associations. Patients who received additional antimicro- bials after initial vancomycin and piperacillin-tazobactam doses were still included in the study population. The effect of additional antimicro- bials, by increasing the spectrum of activity, may have led to improved sepsis treatment. We also did not record the number of vancomycin and piperacillin-tazobactam doses that patients received or the dose ad- ministered, so we were unable to investigate if length and dose of ther- apy was associated with patient outcomes. Finally, we did not include Sepsis-3 definitions in our study; given that many institutions, including ours, still rely upon the older Sepsis-2 definition, we feel the older def- initions are still applicable and used by many providers in the emergen- cy department.

Conclusions

The standard empiric regimen of vancomycin and piperacillin- tazobactam does not provide adequate coverage in approximately a quarter of patients presenting to the ED with severe sepsis or septic shock. Certain populations exist in which alternative or expanded cov- erage should be considered, and that evaluation of prior culture data in individual patients should be an essential aspect of antibiotic selec- tion. In particular, patients transferred to the ED from SNFs, those with COPD, the immunocompromised, those with histories concerning for

C. difficile, and multiple Infection sites may benefit from an alternative initial empiric antibiotic regimen as they are at highest risk for inade- quate coverage with initial vancomycin and piperacillin-tazobactam. Regardless of which combination of broad spectrum antimicrobials an individual Healthcare facility favors-vancomycin with piperacillin- tazobactam or another combination-review of susceptibility patterns should be undertaken at frequent intervals to determine whether an al- ternative or expanded combination of agents should be utilized.

Funding

This research did not receive any specific grant from funding agen- cies in the public, commercial, or not-for-profit sectors.

Conflicts of interest

None.

Author contributions statement

CL, FC, RD, GW, and PF conceived the study and designed the trial. CL, FC, and GW supervised the conduct of the trial. CL, FC, and PF participat- ed in data collection and managed the data, including quality control.

GW and JV provided statistical advice on study design and analyzed the data. CL drafted the manuscript, and all authors contributed sub- stantially to its revision. CL takes responsibility for the paper as a whole.

References

AJEM