Abstract
The objective of this study was to evaluate the accuracy and robustness of a fully automated EUCAST RAST (rapid antimicrobial susceptibility test) directly from positive blood culture and to appreciate its implementation constraints. This study was conducted in two phases: (i) spiked blood culture bottles (BCs) using 779 non-duplicate clinical isolates and (ii) a prospective clinical trial including 534 positive BCs sequentially processed in routine at the Bacteriology Laboratory of Geneva University Hospitals. The RAST results were assessed against EUCAST standardized disk diffusion testing results. Our first finding was that the results of the spiked BCs precisely predicted the clinical trial results. The overall categorical agreements for all species analyzed were greater than 95% at the different time points. RAST for Pseudomonas aeruginosa, however, raised several challenges. The categorical agreement for imipenem was lower than 95% at 6 h and was not improved with longer incubation times. Additionally, piperacillin-tazobactam, ceftazidime, and cefepime cannot be released at 6 h due to suboptimal performances, but the categorical agreement substantially improved at 8 h. Our results establish that the performance of fully automated EUCAST RAST directly from positive blood culture bottles is consistently robust, even for the detection of extended-spectrum β-lactamase (ESBL), carbapenemase-producing bacteria, and methicillin-resistant Staphylococcus aureus (MRSA). The automation markedly enhanced the percentage of readable inhibition zones and reduced the percentage of isolates categorized in the area of technical uncertainty (ATU). In summary, a fully automated EUCAST RAST can substantially improve laboratory workflow by reducing hands-on time and removing the strong constraints linked to manual read-outs at precisely defined times.