Antimicrobial de-escalation (ADE) is a strategy to reduce the spectrum of antimicrobials and aims to prevent the emergence of bacterial resistance. We present a systematic review describing the definitions, determinants and outcomes associated with ADE. We included 2 randomized controlled trials and 12 cohort studies. There was considerable variability in the definition of ADE. It was more frequently performed in patients with broad-spectrum and/or appropriate antimicrobial therapy (P =. 05 to. 002), when more agents were used (P =. 002), and in the absence of multidrug-resistant pathogens (P <. 05). Where investigated, lower or improving severity scores were consistently associated with ADE (P =. 04 to <.001). The pooled effect of ADE on mortality is protective (relative risk, 0.68; 95% confidence interval,. 52-.88). Because the determinants of ADE are markers of clinical improvement and/or of lower risk of treatment failure this effect on mortality cannot be retained as evidence. None of the studies were designed to investigate the effect of ADE on antimicrobial resistance.