1588. Clinical Prediction Tool for Extended-Spectrum Β Lactamase-Producing Enterobacteriaceae as the Etiology of Bacteremia in Solid Organ Transplant Recipients

Abstract

BackgroundMultidrug-resistant Gram-negative bacterial infections have emerged as a significant cause of morbidity among solid-organ transplant (SOT) recipients, leading to challenges in selection of empiric antibiotic therapy. As such, we developed a predictive tool to determine whether these patients are high or low risk for extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae (EB) bacteremia at the time of positive blood culture.MethodsAll SOT recipients admitted to the Hospital of the University of Pennsylvania with blood cultures positive for EB between January 2007 and July 2016 were included in the source cohort. Multivariate logistic regression was used to determine predictors of ESBL-EB bacteremia. Manual forward selection was utilized to maximize area under the receiver operating characteristic curve (AUC), with sequential addition of covariates until achieving an AUC increase of <1%. A scoring system was developed based on the adjusted odds ratios for each covariate. Interval validity was assessed using the Hosmer-Lemeshow statistic, sensitivity, specificity, and the bootstrapping technique.ResultsOver the study period, there were 287 SOT recipients admitted with an EB bacteremia, of which 51 were due to an ESBL-producing organism. The final model that predicted presentation with ESBL-EB bacteremia included: identification of an ESBL-EB on culture in the previous 12 months, trimethoprim-sulfamethoxazole exposure in the prior 6 months, receipt of a kidney transplant, and fluoroquinolone exposure in the prior 6 months. This model achieved an AUC of 0.85, sensitivity of 54.9%, specificity of 96.6%, and Hosmer–Lemeshow statistic of 0.26 (Figure 1). Based on the adjusted odds ratios for each factor, scores of +3, +1, −1, and +1 were assigned to each covariate, respectively. With this scoring system, the maximum attainable score was 5, with any score of ≥3 deemed to be predictive of ESBL-EB bacteremia. After bootstrapping, the adjusted AUC remained at 0.85.ConclusionThe presence of ESBL-EB organism as the etiology of bacteremia in SOT recipients can be predicted using the above scoring system. This tool will inform judicious empiric antibiotic use in this population at the time of positive blood culture. Disclosures All authors: No reported disclosures.