Figure 6

Impact of de-escalation on final outcome. Kaplan Meier survival analysis was conducted comparing patients for which antimicrobials were de-escalated by the attending physicians (Yes, solid line) with those for which antimicrobials were not de-escalated by the attending physicians (No, dashed line) separately for the study period 2006–2009 (panel A) and for the study period 2010–2013 (panel C). P values refer to comparisons between patients under de-escalation and patients without de-escalation. Panels B (study period 2006–2009) and D (study period 2010–2013) refer to forward Cox regression analysis of the impact of de-escalation on the final outcome after adjustment for the presence of severe sepsis/shock, age, gender and for comorbid illnesses. CI: confidence intervals; HR: hazard ratio.