Clinical consequences of very major errors with semi-automated testing systems for antimicrobial susceptibility of carbapenem-resistant Enterobacterales

Objectives In this study we investigated the rate of susceptibility testing discrepancies between semi-automated and reference systems with carbapenem-resistant Enterobacterales (CRE) and the impact of alleged errors by semi-automated systems on guiding targeted therapy for CRE bloodstream infection (BSI). Methods This was a multicentre, retrospective study enrolling patients with monomicrobial BSI caused by CRE from January 2013 to December 2016. Nonduplicate isolates from index blood cultures tested locally with semi-automated systems were centralized at a referral laboratory and retested with a reference broth microdilution or agar dilution method. Results We enrolled 366 patients with CRE-BSI; 220 (60%) were male, and the median age was 67 years (interquartile range, 54–76 years). When compared with the results of the reference methods, those of the semi-automated systems exhibited variable rates of very major errors (VMEs; i.e. false susceptibilities) and major errors (MEs; i.e. false resistances). The highest rates of VMEs were observed with fosfomycin (14%) and colistin (13.9%), and the highest rates of MEs were observed with gentamicin (21%), fosfomycin (7.7%), and tigecycline (34%). Overall, VMEs and MEs led clinicians to prescribe or confirm ineffective therapy in 25 of 341 patients (7%). Receipt of ineffective therapy supported by a misleading susceptibility test was associated with higher 30-day mortality rates by Kaplan–Meier survival curves rates compared with receipt of active therapy (56% vs. 26%; p = 0.002), and the difference was confirmed after adjustment for confounders in a Cox regression model (adjusted hazard ratio: 2.91; 95% CI, 1.62–5.22; p < 0.001). Discussion MEs and VMEs were relatively common with semi-automated susceptibility testing systems. VMEs were associated with inappropriate use of antibiotics and poorer outcomes.