2790. Respiratory Viral Testing and Antimicrobial De-escalation Among Hospitalized Patients at a Tertiary Care Facility, 2015–2016: A Matched Cohort Study Series

Abstract

BackgroundThe use of multiplex respiratory viral panels (RVP) is increasing. They have the potential to reduce unnecessary antibiotic use, but data are limited on their clinical effectiveness. Our objective was to estimate risk differences for antimicrobial de-escalation (discontinuation, intravenous to oral, or spectrum narrowing) between different sequences and results of RVP and rapid polymerase chain reaction (PCR) tests for influenza +/− respiratory syncytial virus.MethodsWe conducted a retrospective chart review of adults (age ≥18 years) admitted to a floor or stepdown unit at University of North Carolina Hospitals who had a respiratory viral test (RVT) within 48 hours of admission between September 2015 and April 2016. We estimated 3-day RDs for the relation between RVT and antimicrobial de-escalation. To control confounding and account for the 37-hour mean lag between PCR (faster) and RVP (slower) tests resulting, we leveraged the treatment decision design over a series of 1:1 matched cohort studies. Each targeted a clinically relevant scenario: (1) ordering RVP test (vs no RVP order) after learning PCR status; (2) learning RVP+ result (vs. no RVP result) after knowing PCR status; (3) learning RVP+ result (vs. RVP–) after knowing PCR–status; and (4) learning RVP+ result (vs. RVP–) given no prior PCR. For each subcohort, referent patients were matched to index patients by race, gender, RVT in prior month (y/n), age (±10 years), and season (±1.7 months).ResultsThe overall cohort (n = 1,342) was 61% White, 29% African American, and 51% female. Median age was 56 years (IQR 39–69). Across all matched subcohorts (Figure 1), the matching success rate was 79–88% and referent frequency of antimicrobial de-escalation ranged 0.6%–1.9%. In scenario 1, ordering RVP results was associated with higher de-escalation (3-day RD 7.6%; 95% confidence intervals [CI] 3.2%, 12.1%). In scenarios 2–4, learning RVP+ results was associated with more frequent de-escalation (3-day RDs 14.8%, 13.8%, and 15.4%).ConclusionRVP testing and positive RVP results were associated with increased antimicrobial de-escalation, although de-escalation was overall infrequent. Future research should assess effect modification across subgroups and evaluate cost-effectiveness. Disclosures All authors: No reported disclosures.