Abstract Rationale Broad-spectrum antibiotics are life-saving medications for critically ill adults. However, prolonged antibiotic use without de-escalation drives antibiotic resistance in Intensive Care Units (ICU). Real-world, longitudinal data on the spectrum of activity (i.e., broadness) and antibiotic consumption are sparse but crucial for informing optimal de-escalation strategies. Therefore, this study characterizes variation in the longitudinal trajectories of antibiotic spectrum coverage among critically ill adults across a diverse set of U.S. healthcare systems. Methods Using a federated approach in 9 U.S. healthcare systems that store data in the Common Longitudinal ICU Format (CLIF), we conducted a retrospective cohort study of critically ill adults admitted to a medical ICU for over 6 hours. Metrics for antimicrobial use included Antibiotic Free Days (AFD), Days of Therapy (DOT), and Days of Antibiotic Spectrum Coverage (DASC). While DOT counts the total number of unique antimicrobials used, DASC assigns a weight to each antibiotic based on its spectrum of activity. To evaluate whether DASC trajectories differed across healthcare systems, we used nested linear models to test for differences in intercepts and slopes (i.e., trajectories) with system and system-by-time interaction terms. Results A total of 501,242 patient days and 130,152 patients (mean age 59.7, 48.2% female, 41.1% non-Hispanic Black) across 41 hospitals were included in the study. The mean SOFA score in the first 24 hours of ICU admission was 4.6 (sd 4.2). Vasopressor and invasive mechanical ventilation were used in 30.7% and 26.6% of patients. ICU mortality was 7.8% (range 5.9 - 14.3%). The mean AFD rate was 46.9%, ranging from 39.1% to 53.9% across healthcare systems. There were 1036.6 DOT per 1000 patient days, but individual healthcare systems ranged from 862.2 to 1395.8 DOT per 1000 patient days. The most commonly prescribed antibiotics were vancomycin, cefepime, and piperacillin-tazobactam with 200.5, 178.2, 121.8 DOT per 1000 patient days, respectively. Overall, DASC decreased from Days 0 to 10 of ICU admission, but baseline (p 0.01) and longitudinal DASC trajectories (p = 0.03) varied between sites (Figure 1). Conclusions For critically ill adults in the medical ICU, vancomycin, cefepime, and piperacillin-tazobactam accounted for the most frequently consumed antibiotics, but baseline and longitudinal trajectories of antibiotic broadness varied across healthcare systems. Further studies should evaluate whether longitudinal antibiotic trajectories for individual patients can inform the optimal timing of de-escalation in critically ill adults. This abstract is funded by: None
Zhang et al. (Fri,) studied this question.