Integrating patient in-hospital transfer patterns into automated outbreak detection systems: a single-center retrospective study

OBJECTIVES: Timely detection of pathogen-related outbreaks in hospitals is important for preventing onward transmission and can be supported by automated outbreak detection systems (AODS). Many methods overlook in-hospital patient transfers and focus only on patient locations at the time of sampling. This study compares three approaches for incorporating patient transfers into AODS. DESIGN: Two existing AODS frameworks, a local percentile-based system and a statistical modeling-based system were extended to include patient transfers: 1) grouping wards into communities based on frequent patient exchange, 2) including prior ward visits in the past 14 days, and 3) including both prior ward visits and time spent on wards. Alerts generated were reviewed for clinical relevance. SETTING: Data from January 2014 to December 2021 from a University Medical Center in the Netherlands. RESULTS: Using the percentile-based approach, the baseline scenario detected 99 possible outbreaks. Extension with ward community groupings, prior ward visits, and prior ward visits accounting for time spent in each ward increased this number with 16 (+15%), 42 (+42%), and 106 (+110%) possible outbreaks, respectively. Of the alerts generated by including individual patient transfer history, 35% were judged as requiring investigation. The trade-off between increased detection and relevance was less favorable for the other approaches. Similar findings were found for statistical modeling-based methods. CONCLUSIONS: Inclusion of patient transfer data in AODS improved sensitivity, at the cost of increasing the alert burden. Therefore, ongoing refinement should further optimize the balance between accurate outbreak detection and a manageable alert burden.