Sixteen studies were included in the review; ten were randomised controlled trials (RCTs). Sample sizes ranged from 21 to 852. The quality of the studies was variable; studies of pharmacological interventions showed higher quality. Three RCTs used an intention-to-treat analysis. Three RCTS blinded participants, clinicians and outcome assessors. Some studies reported very high losses to follow-up.
The overall pooled estimate showed efficacy of interventions in reducing the incidence of delirium (OR 0.64, 95% CI 0.46 to 0.88). The only covariate that appeared to have an impact on the effectiveness of the intervention was the incidence of delirium in the control group; interventions in studies with a rate of 30% or higher in the control group showed larger effect sizes (OR 0.34, 95% CI 0.16 to 0.71; eight RCTs) than those in studies with lower delirium rates (OR 0.76, 95% CI 0.60 to 0.97; eight RCTs).
One-component non-pharmacological interventions did not show a benefit of intervention, but only two studies assessed these and confidence intervals were wide (OR 1.05, 95% CI 0.09 to 11.57)
There was no evidence of publication bias.