Socially assistive robots for people with dementia: Systematic review and meta-analysis of feasibility, acceptability and the effect on cognition, neuropsychiatric symptoms and quality of life.

Background: There is increasing interest in using robots to support dementia care but little consensus on the evidence for their use. The aim of the study is to review evidence about feasibility, acceptability and clinical effectiveness of socially assistive robots used for people with dementia.

Method: We conducted a systematic review and meta-analysis. We searched MEDLINE, EMBASE, PsychINFO, CINHAL, IEEE Xplore Digital Library, and EI Engineering Village from inception to 04 - 02-2022 - included primary studies assessing feasibility, acceptability, or effectiveness of socially assistive robots for people with dementia. Two independent reviewers screened studies for eligibility, and assessed quality. Narrative synthesis prioritized higher quality studies, and random-effect meta-analyses compared robots with usual care (UC) or active control (AC) immediately after the intervention (short-term; ST) or long-term (LT) on cognition, neuropsychiatric symptoms, and quality of life.

Findings: 66 studies and four categories of robots were eligible: Companion robots (Pet and humanoid companion robots), telepresence communication robots, homecare assistive robots and multifunctional robots. PARO (companion robot seal) was feasible and acceptable but limited by its weight, cost, and sound. On meta-analysis, PARO had no ST or LT compared to UC or AC over 5-12 weeks on agitation (ST vs UC, 4 trials, 153 participants: pooled standardized mean difference (SMD) 0.25; - 0.57 to 0.06; LT vs UC; 2 trials, 77 participants, SMD = -0.24; - 0.94, 0.46), cognition (ST vs UC, 3 trials, 128 participants: SMD= 0.03; -0.32, 0.38), overall neuropsychiatric symptoms (ST vs UC, 3 trials, 169 participants: SMD= -0.01; -0.32, 0.29; ST vs AC, 2 trials, 145 participants: SMD =0.02, -0.71, 0.85), apathy (ST vs AC, 2 trials, 81 participants: SMD= 0.14; 0.29, 0.58), depression (ST vs UC, 4 trials, 181 participants; SMD= 0.08; -0.52, 0.69; LT vs UC: 2 trials, 77 participants: SMD =0.01; -0.75, 0.77), anxiety (ST vs UC: 2 trials, 104 participants, SMD= 0.24; -0.85, 1.33) and quality of life (ST vs UC, 2 trials, 127 participants: SMD=-0.05; -0.52, 0.42; ST vs AC: 2 trials, 159 participants, SMD =-0.36, -0.76, 0.05). Robotic animals, humanoid companion robots, telepresence robots and multifunctional robots were feasible and acceptable. However, humanoid companion robots have speech recognition problems, and telepresence robots and multifunctional robots were often difficult to use. There was mixed evidence about the feasibility of homecare robots. There was little evidence on any of these robots' effectiveness.

Conclusion: Although robots were generally feasible and acceptable, there is no clear evidence that people with dementia derive benefit from robots for cognition, neuropsychiatric symptoms, or quality of life. We recommend that future research should use high quality designs to establish evidence of effectiveness.