Background: Physicians make decisions regarding antimicrobial chemotherapy based on clinical and demographic factors, choosing initial empiric therapy without knowing the pathogen or its susceptibilities. Given the various treatment options and resistance mechanisms, treatment of GNR BSI is challenging with 30 day mortality approaching 30%. Using a large cohort of Ec and Kpn BSI, we aimed to characterize empiric antibiotic therapy, comparing treatment before and after Gram stain (GS) results, and summarize clinical outcomes.
Methods: Using a cohort of patients hospitalized within VHA, we used the Corporate Data Warehouse to identify blood cultures positive for Ec or Kpn from 2006 to 2015. We extracted inpatient antimicrobial regimens, demographics, and antibiotic susceptibility testing (AST) results. We excluded cases with missing GS result dates and those not treated with BLs. We defined initial empiric treatment as agents received between specimen collection and GS results; and modified empiric treatment as agents received after GS but before AST results. Patient characteristics, treatments, and outcomes were summarized overall and by organism.
Results: Of 36,531 BSI identified, we analyzed a subset of 21,597 that met our inclusion criteria (Figure). Within this subset of patients, the mean age was 70.3 and all-cause 30-day mortality was 13.9% (2,054 out of 14,735) for Ec and 17.8% (1,220 out of 6,862) for Kpn. Initial empiric treatment included an effective agent in 90.4% (91.2% in Ec, 88.7% in Kpn) of cases. This rate increased to 95.3% (96.0% in Ec, 93.8% in Kpn) for modified empiric treatment. The most commonly prescribed initial empiric BL was piperacillin/tazobactam, observed in 55% of treated patients, followed by ceftriaxone and cefepime in 14% and 11% of treated patients, respectively. Carbapenems were included in 8% of initial and 13% of modified empiric treatments.
Conclusion: In this cohort of older patients with Ec and Kpn BSI, higher rates of effective BL empiric treatment were achieved after GS results. BL empiric regimens consisted mostly of broad spectrum agents. These observations highlight the potential utility of a diagnostic tool available shortly after specimen collection to inform treatment and improve patient outcomes.
S. R. Evans, None
F. Perez, None
H. F. Chambers, None
V. G. Fowler Jr., None
B. N. Kreiswirth, None
R. A. Bonomo, None