Prevalence of Antibiotic-Resistant Pathogens in Culture-Proven Sepsis and Outcomes Associated With Inadequate and Broad-Spectrum Empiric Antibiotic Use

Key Points Question What is the prevalence of antibiotic resistance in community-onset sepsis, and is there risk associated with the receipt of empiric broad-spectrum antibiotics? Findings In this cohort study of 17 430 adults with culture-positive sepsis admitted to 104 US hospitals, 67.0% received empiric broad-spectrum antibiotics, but resistant gram-positive organisms were isolated in only 13.6% of patients and resistant gram-negative organisms in 13.2%. Both undertreatment (failure to cover organisms) and overtreatment (resistant organisms targeted but not isolated) were associated with higher mortality after detailed risk adjustment. Meaning In this study, broad-spectrum antibiotics were frequently administered to patients with community-onset sepsis without resistant organisms, and these therapies were associated with worse outcomes.

Antibiotic susceptibilities were derived from in vitro reports generated by each institution. Individual institutional standards on susceptibility testing were unavailable. Intermediate susceptibilities were treated as resistant. When considering whether antibiotics were active against isolated pathogens, we assessed each potential pathogen-antibiotic combination using reported antibiotic susceptibilities when available. In many cases, however, susceptibilities to specific antibiotics administered were not explicitly listed in susceptibility reports. We therefore created a set of rules to impute these susceptibilities using microbiologic principles and knowledge of the spectrum of activity for each antibiotic-pathogen combination. A "key" to the abbreviations used in these imputations is shown below. In the following sections, the algorithms for each antibioticpathogen imputation are shown.

Imputation Abbrevation
Explanation R This means that an organism will always be considered to be resistant to the antibiotic.  Example: Ceftriaxone for Pseudomonas aeruginosa will always be treated as a non-active antibiotic S This means that an organism will always be considered to be susceptible to the antibiotic.  Example: Ceftriaxone for Methicillin-sensitive Staphylococcus aureus will always be treated as an active antibiotic if missing then E (alternate antibiotic) This means that if the susceptibility to the antibiotic of interest is missing, then the algorithm looks at the antibiotic listed next to "E" and transposes that susceptibility.  Example: For E.coli, "E imipenem" under meropenem means that if meropenem susceptibility is missing, but imipenem is reported as susceptible, then meropenem can be considered to be susceptible too. Some of these have a further logic built it to handle cases where the other antibiotic susceptibility is missing too. These follow a hierarchial order; for example, for "E imipenem; if missing then E ertapenem" then the algorithm first looks to see if there is susceptibility reported for imipenem; if missing, then the algorithm looks for susceptibility for ertapenem. if missing then .
This means that if the susceptibility to the antibiotic is missing, then the algorithm will list that particular antibiotic-organism combination as missing (because an imputation could not be reliably performed). In several cases, this imputation to handle an antibiotic-organism combination as missing comes only if prior imputations were unable to produce a result. Patient encounters that have missing antibioticorganism combinations were excluded from the analysis assessing adequacy of empiric antibiotics or unnecessarily broad antibiotics.* if missing then R This means that the algorithm will report the organism to be resistant to the antibiotic if the susceptibility is missing. if missing then S This means that the algorithm will report the organism to be susceptible to the antibiotic if the susceptibility is missing. Abbreviations: CNS = central nervous system, ICU = intensive care unit, AST = asparate aminotransferase, GCS = Glasgow Coma Scale, Mech Ventilation = mechanical ventilation, Resp Rate = Respiratory Rate, Sys Bld Press = systolic blood pressure, WBC = white blood cell count.

I. GRAM-NEGATIVE ORGANISMS
a Univariate models show the association between each covariate and in-hospital death among all culture-positive community-onset sepsis patients in the cohort (n=17,430). Admission year and hospital characteristics (bed size, region, teaching status) were significant at p<0.001 in the univariate screens but are not shown in the table.
b Multivariable model 1 shows the full model assessing the association between inadequate empiric antibiotic therapy (at least one organism recovered from a clinical culture site that was nonsusceptible to all administered empiric antibiotics) and in-hospital death. This analysis was performed among culture-positive sepsis patients in whom each antibiotic-pathogen combination could be either directly measured or imputed (n=15,183).
c Multivariable model 2 shows the full model assessing the association between unnecessarily broad empiric antibiotic therapy and in-hospital death. This model was done among culturepositive sepsis patients who received adequate empiric antibiotic therapy and in whom each antibiotic-pathogen combination could be either directly measured or imputed (n=12,398). Unnecessarily broad empiric therapy indicates patients who received anti-MRSA (vancomycin, linezolid, daptomycin), anti-VRE (linezolid or daptomycin), anti-Pseudomonal beta-lactams (ceftazidime, cefepime, piperacillin-tazobactam, aztreonam, imipenem, meropenem, doripenem), or anti-ESBL therapy (any carbapenem) when none of those resistant organisms were isolated. Note: the most abnormal (minimum or maximum) values of covariates within the first 2 days of hospitalization were used.