Performance of automated algorithms for detection of clusters of contagious microbial pathogens in hospitals – a scoping review

Abstract

Background When outbreaks of disease-causing pathogens occur in hospitals quick detection is of importance to uncover the source and to prevent infection in other patients. Having automated outbreak detection system applied to hospital microbial data could help IPC-staff in early detection of outbreak. The objective of this thesis was to summarize the literature on the performance of detection algorithms on real microbial hospital data, and to assess the performance of the detection algorithms.

Methods A scoping-review was performed according to guidelines developed by PRISMA. A structured literature search in MEDLINE was performed to find articled published between 2015 and 2025. Eligible articles were screened by reference list and searched in “cited by”-function in PubMed to look for other relevant articles.

Results One randomized control trial, three prospective and ten retrospective studies were summarized in tables and. The results show a high variation of methods for detecting clusters of pathogens. Estimated sensitivity in detecting previously known outbreaks vary between 19% and 100%.

Conclusions Detection algorithms generally perform well in detecting possible outbreaks unnoticed by non-algorithmic detection systems. More prospective studies are needed to assess the utility of detection algorithms.

Background When outbreaks of disease-causing pathogens occur in hospitals quick detection is of importance to uncover the source and to prevent infection in other patients. Having automated outbreak detection system applied to hospital microbial data could help IPC-staff in early detection of outbreak. The objective of this thesis was to summarize the literature on the performance of detection algorithms on real microbial hospital data, and to assess the performance of the detection algorithms.

Methods A scoping-review was performed according to guidelines developed by PRISMA. A structured literature search in MEDLINE was performed to find articled published between 2015 and 2025. Eligible articles were screened by reference list and searched in “cited by”-function in PubMed to look for other relevant articles.

Results One randomized control trial, three prospective and ten retrospective studies were summarized in tables and. The results show a high variation of methods for detecting clusters of pathogens. Estimated sensitivity in detecting previously known outbreaks vary between 19% and 100%.

Conclusions Detection algorithms generally perform well in detecting possible outbreaks unnoticed by non-algorithmic detection systems. More prospective studies are needed to assess the utility of detection algorithms.