562. Electronic Implementation of the Ventilator-Associated Events Algorithm: the Solution for Variability?
Session: Oral Abstract Session: Advancing Public Reporting and Surveillance of HAIs
Thursday, October 3, 2013: 3:00 PM
Room: The Moscone Center: 200-212
Background: Reliable surveillance methods are indispensable for benchmarking of healthcare-associated infection rates. The National Healthcare Safety Network (NHSN) recently introduced surveillance of ventilator-associated events (VAE), including ventilator-associated conditions (VAC). This new algorithm is amenable to automated implementation and strives for more consistent interpretation. We assessed the feasibility and reliability of automated implementation.

Methods: Retrospective analysis of an ICU cohort with prospective assessment of ventilator-associated pneumonia (VAP) in 2 academic medical centers (January 2011 - June 2012).  The algorithm was electronically implemented as specified by NHSN using minute-to-minute ventilator data. Two minor modifications were developed to improve stability and comparability with manual surveillance (10thpercentile & intermittent ventilation). Concordance was assessed between the algorithms and prospective surveillance. Attributable mortality of VAC was estimated by multivariable competing-risk survival analysis.

Results: 2080 patients contributed 2296 episodes of mechanical ventilation (MV). VAC incidence was 10.0/1000 MV days. Prospective surveillance identified 8 VAP cases/1000 MV days.  The original VAC algorithm detected 32% (38/115) of patients affected by VAP; positive predictive value was 25% (38/152). Using the 10th percentile identified the same number of VAC cases, however only 116 were identical. VAC incidence was 24.9/1000 MV days with the intermittent ventilation modification. Concordance between the original algorithm and the modified versions was suboptimal. Estimates of attributable mortality varied by implementation: original VAC subdistribution hazard ratio (sdHR) = 4.33, 10thpercentile sdHR = 6.26 and intermittent ventilation sdHR = 2.40.

Conclusion: Concordance between manual VAP surveillance and the VAE algorithm was poor. Although electronic implementation of the VAE algorithm was feasible, small variations considerably altered the events detected and their effect on mortality. Using the current specifications, comparability across institutions using different electronic or manual implementations remains questionable.

Maaike S.M. van Mourik, MD, MSc1, Peter M.C. Klein Klouwenberg, MD, PharmD1, David S.Y. Ong, MD PharmD1, Janneke Horn, MD PhD2, Marcus J. Schultz, MD PhD2, Olaf L. Cremer, MD PhD3 and Marc Bonten, MD PhD1,4, (1)Department of Medical Microbiology, University Medical Center Utrecht, Utrecht, Netherlands, (2)Department of Intensive Care Medicine, Acadamic Medical Center, University of Amsterdam, Amsterdam, Netherlands, (3)Department of Intensive Care, University Medical Center Utrecht, Utrecht, Netherlands, (4)Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, Netherlands

Disclosures:

M. S. M. van Mourik, None

P. M. C. Klein Klouwenberg, None

D. S. Y. Ong, None

J. Horn, None

M. J. Schultz, None

O. L. Cremer, None

M. Bonten, None

Findings in the abstracts are embargoed until 12:01 a.m. PST, Oct. 2nd with the exception of research findings presented at the IDWeek press conferences.