Outcome of inadequate empirical antibiotic therapy in emergency department patients with community-onset bloodstream infections

Received 2 April 2012; returned 27 May 2012; revised 28 October 2012; accepted 31 October 2012Background: Studies on the effect of inadequate empirical antibiotic therapy on the outcome of patients withsystemic infection have led to inconsistent results.Methods: We analysed data from a comprehensive clinical database collected prospectively in a university hos-pital between 2008 and 2009. All adult patients who registered in the emergency department (ED) with abloodstream infection (BSI) were enrolled. Empirical therapy was considered adequate if it included antimicro-bials to which the specific isolate displayed in vitro susceptibility and that were administered within 24 h of EDadmission. The propensity score (PS) was created by a logistic regression model predicting inadequate empiricaltherapy. PS-adjusted multivariate analysis was performed by the Cox regression model. The Mortality in Emer-gency Department Sepsis (MEDS) score was used for the adjustment of residual confounding due to differencesin the baseline clinical severity of disease.Results: Out of 937 episodes of bacteraemia, 255 (27.2%) patients received inadequate empirical antimicrobialtherapy. A crude analysis showed that inadequate antibiotic therapy was associated with higher mortality rates(hazard ratio 1.78, 95% CI 1.30–2.45). PS-adjusted multivariate analyses also showed a significant adverseimpact (hazard ratio 1.59, 95% CI 1.14–2.28). The clinical disease severity significantly modified the effectof inadequate antibiotic therapy on survival. The magnitude of the adverse impact of inadequate antibiotictherapy decreased with the increasing severity of sepsis (P¼0.009).Conclusions: Inadequate empirical antimicrobial therapy for community-onset BSI was associated with higher30 day mortality rates. Study populations with different clinical severities may have different results, which mayhelp to partly explain the heterogeneous findings in many similar studies.Keywords: mortality, effect modification, propensity score