1231. Legionella Variability From Routine Environmental Testing Across All Veterans Health Administration (VHA) Medical Facilities

Abstract

BackgroundVHA Legionella prevention policy requires quarterly testing of potable water samples, for its 170 medical facilities (“stations”) distributed across the United States. We modeled the variability in Legionella positivity rates by location structure and by time to understand Legionella prevalence and distribution across VHA nationwide. Our goal was to understand when, where and why variations in Legionella positivity happens across VHA facilities.MethodsData from quarterly water samples from sinks and showers from 2015 through 2017 and for which complete information was reported were used for the model. A multi-level Bayesian logistic regression model was run in R version 3.5.1. The hierarchical location group levels consisted of room nested within floor, within building, within station, within region. The time group-level effects included quarter nested within year. Variabilities within groups were estimated as standard deviation (SD) on the log-odds scale.ResultsAmong 138,553 samples, there was little seasonal effect (SD: 0.32) in Legionella positivity based on the quarter in which they were sampled. The largest variability in Legionella positivity occurred at the station level (SD: 2.38), with substantial variation at the building level also (SD: 1.85). The 5% of stations most likely to be positive for Legionella represented only 7.5% of total samples but accounted for 39.7% of all positive samples. The 5% of stations least likely to be positive for Legionella represented 10.4% of total samples, but only had 2 positive samples.ConclusionBuildings with the highest probability for Legionella positivity are clustered together within stations. We saw no major seasonal variations in Legionella positivity across facilities. We were able to better predict stations with higher positivity as well as lower overall positivity for Legionella water sampling. The observed dominant station-level effects could be due to overarching influences such as a single water source and suggests approaches at this level can impact Legionella control. These results demonstrate a mechanism for understanding the distribution and probability of Legionella and can inform prevention practices and future policy.Disclosures All authors: No reported disclosures.