1800. Clinical Impact of Real-Time Predictive Model to Facilitate Antibiotic Prescribing in Gram-Negative Bacteremia

Abstract

BackgroundDelay in effective antibiotic administration in severe infections such as bacteremia is associated with worse clinical outcomes. We implemented previously validated software that uses real-time predictive modeling to determine patient-specific antibiograms (PS-ABG). The software allowed prescribers to run the model on their individual patients. It also automatically evaluated positive blood cultures, alerting the antibiotic stewardship team if there was <90% chance of the organism being susceptible to current antibiotic therapy.MethodsWe performed a quasi-experimental study to evaluate clinical outcomes in patients with Gram-negative rod (GNR) bacteremia 18 months before (PRE) and 6 months after (POST) implementation of the software. Primary outcome was median time to effective antibiotic. Secondary outcomes included in-hospital mortality, utilization of antibiotics used for multidrug-resistant GNRs (MDR-GNR), median time to effective antibiotic in organisms resistant to at least one first-line antibiotic for sepsis, and length of stay.ResultsThe change per month in the primary outcome did not differ between the PRE and POST periods (P = 0.48) (figure). Time to effective antibiotics in GNR bloodstream infections that were resistant to at least one first-line antibiotic for sepsis (cefepime, piperacillin–tazobactam, or levofloxacin) was lower following the intervention (15.8 hours vs. 13.7 hours, P = 0.11), and mortality decreased following the intervention (14.6% vs. 10.0%, P = 0.11) although these differences were not statistically significant. There was no difference in other secondary outcomes between PRE and POST groups: length of stay (7.7 vs. 7.5 days, P = 0.74) and days of therapy of MDR-GNR agents per 30 days of hospitalization (3.5 vs. 2.5, P = 0.09).ConclusionThere was no difference in median time to effective antibiotic in all patients with GNR bacteremia. There was lower in-hospital mortality in the POST group and shorter time to effective antibiotic therapy in GNR bacteremia resistant to at least one first-line antibiotic for sepsis, although these differences were not statistically significant. Additional study in larger cohorts over longer periods is warranted to determine whether PS-ABGs improve clinical outcomes in patients with more resistant GNR bacteremia. Disclosures All authors: No reported disclosures.