Abstract
BACKGROUND: Multidrug-resistant organisms (MDROs) have become a serious public health concern worldwide. Screening for MDROs can encompass various phenotypic and genotypic approaches. Recently, culture-based MDROs screening has substantially benefitted from laboratory automation and artificial intelligence. OBJECTIVES: To evaluate the performance of PhenoMATRIX® PLUS for an institution-wide culture-based rectal screening for CPOs, ESBL-E, Acinetobacter, and VRE carriage. METHODS: The number of non-duplicated rectal Eswabs specimens used to generate culture plate images for the PhenoMATRIX® PLUS machine learning phase amounted to 2'900 for chromID™ ESBL agar, 2'403 for chromID™ OXA-48 agar, 2'031 for CHROMagar™ Acinetobacter agar, and 9'333 for chromID™ VRE agar. All such plates and images were classified by manual reading as either positive or negative. The validation of the settings derived during the learning phase was achieved on an additional 4'517 non-duplicate specimens. Digital images of 22'585 media plates (i.e., five per specimen) were prospectively analyzed by PhenoMATRIX® PLUS. All media plates were incubated in the WASPLab® and imaged at predefined time points. RESULTS: According to the manual work-up results, the output agreements of the PhenoMATRIX® PLUS for the three results "no growth," "negative," and "Send to Reader" reached 98.5, 99.6%. and 74.8%, respectively. Importantly, no false negative results were returned by PhenoMATRIX® PLUS in the present study. CONCLUSION: PhenoMATRIX® PLUS contributes to a significant reduction in the workload. Therefore, technologists' routine tasks have evolved to the confirmatory tests of the presumptively positive colonies, allowing better management of the laboratory staff and the resources.