Table 2.

High-Impact or Notable Outbreaks Detected by EDS-HAT

Outbreak	Details
Vancomycin-resistant Enterococcus faecium outbreak associated with IR and injection of sterile contrast [6]	This outbreak involved 10 initial patients and was ongoing when it was discovered. The EDS-HAT ML algorithm identified IR as a significant transmission route (OR: 43.8; P < .01; 95% CI: 5.6 to 346). Nine patients, including 3 with bacteremia, were identified as having IR procedures involving unsterile practices in the preparation of contrast. Safe practices and enhanced environmental cleaning were implemented and no additional IR-associated infections occurred. Subsequently, transmission of the outbreak strain occurred among 4 patients on shared hospital units.
Pseudomonas aeruginosa outbreak associated with gastroscopy [5]	This outbreak comprised 6 patients housed on different units over 7 months. Two patients had bacteremia, 3 had pneumonia, and 1 had a urinary tract infection. The EDS-HAT ML algorithm detected gastroscopy as a significant route for 4 patients (OR: 300.6; P < .01; 95% CI: 15.8 to 5690.5) with a fifth patient who did not have a charge code that reflected the gastroscopy procedure but who had a clinical note reflecting the procedure that was identified on manual EHR review. A post-disinfection gastroscope culture performed as part of routine IP practice was positive for P. aeruginosa; the isolate was sequenced and belonged to the outbreak, confirming gastroscopy as the responsible transmission route.
Outbreaks of multiple pathogens at the embedded chronic care facility	EDS-HAT ML identified 11 clusters involving 38 patients over 22 months, with a range of 2–9 total patients per cluster; 25 (65.8%) patients had this facility as a plausible transmission route. Pathogens included C. difficile (6 clusters), Klebsiella pneumoniae (1 cluster), MRSA (1 cluster), P. aeruginosa (2 clusters), and VRE (1 cluster). Three patients with C. difficile in 3 clusters were subsequently transferred to our institution and had unit-based commonalities with 3 additional patients who later developed C. difficile infection suggesting continuing transmission.
Outbreaks of multiple pathogens on an ICU	There were 12 clusters with 57 patients (range: 2–14), of whom 28 (49.1%) had a single ICU stay identified by EDS-HAT ML as the potential transmission route. Organisms included C. difficile (3 clusters involving 10 patients), K. pneumoniae (3 clusters involving 16 patients), P. aeruginosa (1 cluster involving 3 patients), Serratia marcescens (1 cluster involving 2 patients), and VRE (4 clusters involving 26 patients).
C. difficile outbreaks associated with wound care	There were 9 C. difficile clusters, ranging in size from 2 to 12 patients. Of 52 patients, 29 (55.8%) had wound care service identified as a potential transmission route, with exposures occurring 1–92 days (mean: 16 days; median: 9 days) before the positive test for C. difficile. This consult service involved nurses providing management of sacral pressure ulcer wounds.
MRSA infections associated with EEG	This cluster consisted of 2 patients with culture dates separated by 8 days. The EDS-HAT ML algorithm identified EEG as a transmission route. Manual EHR review determined that both patients had a bedside EEG performed on the same day on separate units by the same physician and technician, 2 and 10 days before positive culture dates.

Abbreviations: CI, confidence interval; EDS-HAT, Enhanced Detection System for Healthcare-Associated Transmission; EEG, electroencephalography; EHR, electronic health record; ICU, intensive care unit; IP, infection prevention; IR, interventional radiology; ML, machine learning; MRSA, methicillin-resistant Staphylococcus aureus; OR, odds ratio; VRE, vancomycin-resistant Enterococcus faecium.