BACKGROUND
Hospital outbreaks of antimicrobial-resistant (AMR) bacteria should be detected and controlled as early as possible.
AIM
To develop a framework for automatic detection of AMR outbreaks in hospitals.
METHODS
Japan Nosocomial Infections Surveillance (JANIS) is one of the largest national AMR surveillance systems in the world. For this study, all bacterial data in the JANIS database were extracted between 2011 and 2016. WHONET, a free software for the management of microbiology data, and SaTScan, a free cluster detection tool embedded in WHONET, were used to analyse 2015-2016 data of eligible hospitals. Manual evaluation and validation of 10 representative hospitals around Japan were then performed using 2011-2016 data.
FINDINGS
Data from 1,031 hospitals were studied; mid-sized (200-499 bed) hospitals accounted for 60%, followed by large (≥500 beds) hospitals (24%) and small (<200 beds) hospitals (16%). Large hospitals resulted in more clusters detected. Most of the clusters included five or fewer patients. From the in-depth analysis of 10 hospitals, approximately 80% of the detected clusters were unrecognised by infection control staff because the bacterial species involved were not included in the priority pathogen list for routine surveillance. In particular, in two hospitals, clusters of more susceptible isolates were detected before outbreaks of more resistant pathogens.
CONCLUSION
WHONET-SaTScan can automatically detect clusters of epidemiologically-related patients based on isolate resistance profiles beyond lists of high-priority AMR pathogens. If clusters of more susceptible isolates can be detected, it may allow early intervention in infection control practices before outbreaks of more resistant pathogens occur.