637. Automated Outbreak Detection of Hospital-Associated Infections
Session: Oral Abstract Session: HAI Reporting: The Devil is in the Details
Thursday, October 8, 2015: 3:15 PM
Room: 5--AB
Background: Containment of hospital outbreaks requires timely identification, investigation, and response. Current detection methods rely on empirical rules covering temporal or spatial clustering of a limited number of prespecified pathogens.

Methods: We applied a multi-dimensional scan statistic (SaTScan), accessed through WHONET software, to 83 years of historical microbiology data from 44 hospitals using Premier SafetySurveillor. We identified clusters of cultures yielding pathogenic species, where clusters were defined by antimicrobial resistance profiles, locations, and specialty services. We included clusters involving ≥3 patients and occurring by chance less than once per year. For comparison, a convenience sample of hospitals provided clusters identified using their usual methods. Infection preventionists (IPs) at each hospital rated the value of the information about each statistically-based cluster and the automated outbreak detection system. 

Results: We identified 222 clusters in 44 hospitals. The number of clusters ranged from 0-4 per 100 beds/year. The majority were identified based on resistance patterns, others by common ward or service. The organisms causing 135 clusters were not under routine surveillance. Ten hospitals provided data on known clusters, and only 20% (10/51) of those clusters were statistically affirmed. The outbreak algorithm detected 69 clusters that were not known to the hospitals. IPs returned surveys on 153 of the 222 clusters. They would have wanted notification about 81% of the clusters and were moderately or highly concerned about 35%. Half of the clusters (51%) expanded after detection. If these had been detected in real time, it would have been possible to intervene and possibly curtail the outbreak. Eight of 9 clusters identified by both the software and hospitals were detected earlier by the software (mean difference 5.5 days). All IPs felt that an automated outbreak detection tool would improve their ability to detect outbreaks (80% moderate or large extent) and streamline their work (65% moderate or large extent).

Conclusion: Automated outbreak detection can increase the efficiency and scope of identification of nosocomial clusters, including pathogens not under routine surveillance.

Meghan Baker, MD, ScD1, Deborah S. Yokoe, MD, MPH, FIDSA, FSHEA2, John Stelling, MD, MPH3, Rebecca E. Kaganov, BA1, Alyssa R. Letourneau, MD, MPH4, Thomas O'brien, MD5, Martin Kulldorff, PhD6, Damilola Babalola, MD, MPH7, Craig Barrett, Pharm.D, BCPS8, Marci Drees, MD, MS9, Richard Platt, MD, MS, FIDSA, FSHEA6, Susan S. Huang, MD, MPH, FIDSA, FSHEA10 and For the CDC Prevention Epicenters Program, (1)Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA, (2)Medicine, Brigham and Women's Hospital, Boston, MA, (3)Brigham and Women's Hospital, Boston, MA, (4)Infectious Diseases, Massachusetts General Hospital, Harvard Medical School, Boston, MA, (5)Brigham & Women's Hospital, Boston, MA, (6)Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA, (7)Brigham and Women’s Hospital, Division of Infectious Diseases, Department of Medicine, Boston, MA, (8)Premier, Inc., Charlotte, NC, (9)Christiana Care Health System, Newark, DE, (10)Division of Infectious Diseases and Health Policy Research Institute, University of California Irvine School of Medicine, Irvine, CA

Disclosures:

M. Baker, None

D. S. Yokoe, None

J. Stelling, None

R. E. Kaganov, None

A. R. Letourneau, None

T. O'brien, None

M. Kulldorff, None

D. Babalola, None

C. Barrett, Premier, Inc.: Employee , Research grant and Salary

M. Drees, None

R. Platt, None

S. S. Huang, Sage Inc.: Investigator , Contributed product for ABATE Study
Molnlycke: Investigator , Contributed product for ABATE Study
3M: Investigator , Conducting studies in which participating healthcare facilities are receiving product from 3M.

<< Previous Abstract | Next Abstract

Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 7th with the exception of research findings presented at the IDWeek press conferences.