De-escalation of antimicrobial therapy in ICU settings with high prevalence of multidrug-resistant bacteria: a multicentre prospective observational cohort study in patients with sepsis or septic shock.

Background: De-escalation of empirical antimicrobial therapy, a key component of antibiotic stewardship, is considered difficult in ICUs with high rates of antimicrobial resistance. Objectives: To assess the feasibility and the impact of antimicrobial de-escalation in ICUs with high rates of antimicrobial resistance. Methods: Multicentre, prospective, observational study in septic patients with documented infections. Patients in whom de-escalation was applied were compared with patients without de-escalation by the use of a propensity score matching by SOFA score on the day of de-escalation initiation. Results: A total of 262 patients (mean age 62.2 +/- 15.1 years) were included. Antibiotic-resistant pathogens comprised 62.9%, classified as MDR (12.5%), extensively drug-resistant (49%) and pandrug-resistant (1.2%). In 97 (37%) patients de-escalation was judged not feasible in view of the antibiotic susceptibility results. Of the remaining 165 patients, judged as patients with de-escalation possibility, de-escalation was applied in 60 (22.9%). These were matched to an equal number of patients without de-escalation. In this subset of 120 patients, de-escalation compared with no de-escalation was associated with Lower all-cause 28 day mortality (13.3% versus 36.7%, OR 0.27, 95% CI 0.11-0.66, P= 0.006); ICU and hospital mortality were also Lower. Deescalation was associated with a subsequent collateral decrease in the SOFA score. Cox multivariate regression analysis revealed de-escalation as a significant factor for 28 day survival (HR 0.31, 95% CI 0.14-0.70, P= 0.005). Conclusions: In ICUs with high Levels of antimicrobial resistance, feasibility of antimicrobial de-escalation was Limited because of the multi-resistant pathogens isolated. However, when de-escalation was feasible and applied, it was associated with Lower mortality.