Epidemiology of Inappropriate Empiric Antibiotic Therapy for Bacteremia Based on Discordant In vitro Susceptibilities: Risk factors and Taxon-level Variation in Burden and Outcome in 156 US hospitals, 2000–2014

Background Discordance between in vitro susceptibility and empiric antibiotic therapy is inextricably linked to antibiotic resistance and decreased survival in bloodstream infections (BSI). However, its prevalence, patient- and hospital-level risk factors, and impact on outcome in a large cohort and across different pathogens remain unclear. Methods We examined in vitro susceptibility interpretations for bacterial BSI and corresponding antibiotic therapy among inpatient encounters across 156 hospitals from 2000 to 2014 in the Cerner Healthfacts database. Discordance was defined as nonsusceptibility to initial therapy administered from 2 days before pathogen isolation to 1 day before final susceptibility reporting. Discordance prevalence was compared across taxa; risk factors and its association with in-hospital mortality were evaluated by logistic regression. Adjusted odds ratios (aOR) were estimated for pathogen-, patient- and facility-level factors. Results Of 33,161 unique encounters with BSIs, 4,219 (13%) at 123 hospitals met criteria for discordant antibiotic therapy, ranging from 3% for pneumococci to 55% for E. faecium. Discordance was higher in recent years (2010–2014 vs. 2005–2009) and was associated with older age, lower baseline SOFA score, urinary (vs. abdominal) source and hospital-onset BSI, as well as ≥500-bed, Midwestern, non-teaching, and rural hospitals. Discordant antibiotic therapy increased the risk of death [aOR = 1.3 [95% CI 1.1–1.4]). Among Gram-negative taxa, discordant therapy increased risk of mortality associated with Enterobacteriaceae (aOR = 1.3 [1.0–1.6]) and non-fermenters (aOR = 1.7 [1.1–2.5]). Among Gram-positive taxa, risk of mortality from discordant therapy was significantly higher for S. aureus (aOR = 1.3 [1.1–1.6]) but unchanged for streptococcal or enterococcal BSIs. Conclusion The prevalence of discordant antibiotic therapy displayed extensive taxon-level variability and was associated with patient and institutional factors. Discordance detrimentally impacted survival in Gram-negative and S. aureus BSIs. Understanding reasons behind observed differences in discordance risk and their impact on outcomes could inform stewardship efforts and guidelines for empiric therapy in sepsis. Disclosures All authors: No reported disclosures.