Intelligent Robot Interventions for People With Dementia: Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background The application of intelligent robots in therapy is becoming more and more important for people with dementia. More extensive research is still needed to evaluate its impact on behavioral and psychological dementia symptoms, as well as quality of life in different care settings. Objective The purpose of this research is to methodically assess how well intelligence robot interventions work for patients with dementia. Methods In accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines, a comprehensive search was conducted on PubMed, CINAHL, the Cochrane Library, Embase, and Web of Science from the time of their founding to February 2024, to identify relevant randomized controlled trials on the use of intelligent robots in people with dementia. Two authors (WF and RZ) independently applied the Cochrane Collaboration bias assessment tool to assess the included studies’ quality. The intervention effect of intelligent robots on patients with dementia was summarized using a fixed-effect model or a random-effects model with Stata software (version 16.0; StataCorp). Subgroup analysis was performed according to the intelligent robot type and the intervention duration. Publication bias was tested using funnel plots, Egger tests, and the trim-and-fill method. Results In total, 15 studies were finally included for systematic review, encompassing 705 participants, of which 12 studies were subjected to meta-analysis. The meta-analysis found that compared with the control group, intelligent robot intervention significantly reduced the levels of agitation (standardized mean difference –0.36, 95% CI –0.56 to –0.17; P<.001) and anxiety (weighted mean difference –1.93, 95% CI –3.13 to –0.72; P=.002) in patients with dementia. However, the intervention of intelligent robots had no significant effect on the following (all P>.05): cognitive function, neuropsychiatric symptoms, depression, quality of life, step count during the day, and the hours of lying down during the night of patients with dementia. Subgroup analysis revealed that the improvement of depression was related to the duration of the intervention (≤12 vs 12 weeks: 0.08, 95% CI –0.20 to 0.37 vs –0.68, 95% CI –1.00 to –0.37; P=.26) and was independent of the type of intelligent robots (animal robots vs humanoid robots: –0.30, 95% CI –0.75 to 0.15 vs 0.07, 95% CI –0.21 to –0.34; P=.26). Conclusions This study shows that intelligent robot intervention can help improve the agitation and anxiety levels of people with dementia. The intervention may be more effective the longer it is implemented. The appearance of the intelligent robot has no effect on the intervention effect. Further research is needed to help collect physiological data, such as physical activity in people with dementia; explore the impact of other intelligent robot design features on the intervention effect; and provide a reference for improving intelligent robots and intervention programs. Trial Registration PROSPERO CRD42024523007; https://tinyurl.com/mwscn985