LAUSR

Abstract Background Reduced gastrointestinal tract bacterial community diversity has been associated with increased risk for healthcare-associated infections, including Clostridium difficile infection. We sought to develop a model for concomitant change in bacterial community diversity at gastrointestinal and respiratory tract sites, drawing upon a recently completed cohort study of 92 subjects recruited from a long-term acute care hospital (LTACH) for dense longitudinal oral, endotracheal aspirate (ET), and stool specimen collection. Methods We evaluated the first 30 subjects enrolled from the LTACH cohort, for whom complete antibiotic administration data and 16S rRNA gene (V1–V2 amplicon) sequencing data were available. Sequencing was performed via the Illumina HiSeq platform; operational taxonomic units (OTUs) were formed and taxonomy assigned (GreenGenes 13.8) via the QIIME 1.9.1 pipeline. Generalized linear mixed effects models were fit using R (3.5.0), Stan (2.1.7), via the “rstan” and “rethinking” packages. Results We evaluated 472 subject-days of study enrollment across the 30 subjects (median 15 days/subject). ET specimens were available for all subject-days; oral and stool for 357 and 177 subject-days, respectively. We modeled daily change in Shannon diversity across oral, ET, and stool specimens, parameterized with daily exposure to cefepime, piperacillin–tazobactam, meropenem, IV vancomycin, and oral vancomycin. All parameters fit with Rhat value lower than 1.1. Absent antibiotic exposure on the previous day, the daily change in Shannon diversity at all sites was near zero. The largest observed effect was oral vancomycin on stool (daily delta Shannon: −0.6, 95% CI: −1.38 to 0.09). All estimated effects for intravenous antibiotics on the stool, and for all antibiotics at other sites were smaller. Conclusion Small daily changes in bacterial community diversity were attributable to individual antibiotics, but all 95% certainty intervals crossed zero in this pilot study. Further work will focus on modeling specific taxonomic changes attributable to individual antibiotics and antibiotic interactions. Disclosures All authors: No reported disclosures.