Four studies (3,074 participants) were included in the review: one RCT (1,133 participants), one non-RCT (139 participants) and two controlled before-and-after studies (1,802 participants). Low risks of bias were mainly assessed for allocation concealment (three studies), healthcare outcomes reported at baseline (three studies) and protection against contamination (three studies). High or unknown risks of bias were assessed for sequence generation, reporting of baseline characteristics and completeness of data/blinding.
At most 36% of intervention participants and 41% of control participants were judged as having the capacity to make decisions (three studies). Increased documentation of patient preferences for care was observed in the intervention groups compared with controls (three studies); two studies reported that this difference was significant.
Compared with control groups, intervention groups showed significant reductions in hospitalisation rates (two studies) and a significantly increased rate of hospice service utilisation (one study). Only one study investigated participant health as an outcome; it reported a significant increase in uptake of pain assessment for an intervention group over the control group. No significant difference in pain medication use was found between them.