2397. Comparing Predictive Performance of INCREMENT Scores on Mortality Among Patients With Carbapenem-Non-Susceptible (CNS) Klebsiella pneumoniae (Kp) and Enterobacter cloacae Complex (Ecc) Bloodstream Infections (BSI) in the Veterans Health Administration (VHA)

BACKGROUND: INCREMENT is an international collaborative study of BSI caused by extended-spectrum β-lactamase (ESBL) or carbapenemase-producing Enterobacteriaceae (CPE) that has developed and validated predictive models for mortality. Most CNS Enterobacteriaceae BSI in the VHA are either Klebsiella pneumoniae (Kp) or Enterobacter cloacae complex (Ecc). We applied the INCREMENT score for CPE to predict mortality in patients with CNS-Kp and CNS-Ecc BSIs in the VHA and compared the distribution and predictive performance of the score across organisms. METHODS: Using nationwide VHA databases, unique patients in the continental United States with Kp or Ecc BSI post 48 hours of hospitalization from 2006 to 2015 were identified. Isolates with intermediate susceptibility or resistance to any tested carbapenem were considered non-susceptible. We used databases and medical records to obtain clinical characteristics, treatment, and outcomes, and applied INCREMENT criteria and definitions to calculate a prediction score. We compared the distribution of the scores by organism and used receiver operating curve methods to compare predictive performance between Kp and Ecc BSI. RESULTS: We identified 57 patients with CNS-Ent and 140 with CNS-Kp BSI. The demographics and infection characteristics were highly consistent across organisms, both afflicting patients who were predominantly male, older and chronically ill. Mortality at 14 days was 39% in CNS-Ecc and 38% in CNS-Kp. Similar proportions (65% of Ecc and 68% of Kp) met the criteria for an INCREMENT score: monomicrobial and alive over 48 hours after culture specimen. The distribution of scores was similar within mortality outcomes across organisms, with the highest scores observed in Kp patients who died (Figure 1). The ROC areas under the curve were 0.71 for CNS-Ecc and 0.75 for CNS-Kp (Figure 2). A multivariable logistic model predicting mortality detected neither an organism effect nor an interaction of organism and INCREMENT score. CONCLUSION: The INCREMENT score, validated in a CPE cohort predominantly comprised of Kp, performed similarly well across CNS-Ent and CNS-Kp patients in our cohort. This suggests the model is robust to CNS organisms of undetermined resistance mechanism and that the association between INCREMENT and mortality is consistent across Kp and Ecc. [Image: see text] [Image: see text] DISCLOSURES: All authors: No reported disclosures.