Background: In gram-negative bacteremia (GNB), administrative data suggest that difficult-to-treat resistance (DTR; i.e co-resistance to all first-line antibiotics) increases mortality. However, adequate risk-adjustment for severity of illness (SOI) may require granular laboratory and physiologic data.
Methods: Adult inpatients with GNB were identified from electronic health records (EHRs) of 140 hospitals in the Cerner Healthfacts database between 2009 and 2015. Mortality from DTR (intermediate/resistant in vitro to β-lactams including carbapenems and fluoroquinolones) was compared to GNB phenotypes susceptible to at least one first-line agent, but otherwise resistant to carbapenems (CR), extended-spectrum cephalosporins (ECR), or fluoroquinolones (FQR) per US Centers for Disease Control and Prevention surveillance definitions. Relative risk of mortality was adjusted (aRR) for age, sex, baseline Sequential Organ Failure Assessment (SOFA) score, Elixhauser comorbidity index, GNB source, taxon, hospital vs. community onset, year, and hospital region, bed capacity, and urban and teaching status using Poisson regression.
Results: Of 25,448 unique GNB encounters, 207 (1%) met DTR criteria. DTR patients were 2-fold more likely to receive intravenous colistin and 5-fold more likely to receive tigecycline compared to CR cases susceptible to ≥1 first line agent. Crude mortality varied considerably by taxon and resistance phenotype, but resistance per se was associated with only a minority of overall deaths (DTR=3% of deaths; any of the four resistance phenotypes=28% of deaths; Fig 1). Inclusion of EHR-derived, baseline SOFA scores in SOI adjustments decreased aRR effect estimates; nonetheless, all resistance phenotypes still significantly increased mortality (Fig 2A). Among resistance phenotypes, aRR of mortality was similar for DTR vs. CR (aRR=1.18; 95% CI, 0.91-1.54; p=0.2), but higher for DTR vs. ECR (aRR=1.26 [1.01-1.58]; p=0.04), and DTR vs. FQR (aRR=1.36 [1.08-1.70]; p=0.008), respectively (Fig 2B).
Conclusion: DTR is associated with non-survival and greater use of reserve antibiotics in GNB, but adds little to the risk of death associated with CR. The impact of resistance on survival is attenuated but still present even after risk adjustment using granular clinical data.
S. S. Kadri,
E. E. Ricotta, None
J. Strich, None
A. Babiker, None
C. Rhee, None
M. Klompas, None
J. P. Dekker, None
J. H. Powers III, None
R. L. Danner, None
J. Adjemian, None
See more of: Poster Abstract Session