Forest plot demonstrating interactions between sepsis and the effect of sedative group on delirium/coma-free days, delirium-free days, coma-free days, and ventilator-free days. For each outcome, the adjusted difference in the means between the dexmedetomidine group and lorazepam group is presented, first for the septic patients (heavy circle) and then for the non-septic patients (heavy triangle), along with 95% confidence intervals (CI) for the difference. Differences, CIs and P values were calculated using bootstrap multiple linear regression, adjusting for age, the acute physiology component of the Acute Physiology and Chronic Health Evaluation (APACHE) II score at enrollment, administration of drotrecogin alfa (activated), treatment group, sepsis, and treatment group by sepsis interaction. If the difference in means is greater than 0, it reflects an improved outcome with dexmedetomidine; if less than 0, then patients on lorazepam had a better outcome. We considered a P value for interaction less than 0.15 to indicate that the effect of sedative group on the outcome in question was different for septic patients than for non-septic patients. A P value for interaction of 0.15 or more, alternatively, indicated that the effect of sedation group on outcomes was the same for all patients, regardless of sepsis.