Results of literature search
Two hundred fifty-two papers were obtained through literature search by using search terms. 160 similar and duplicate papers were excluded. 92 distinct articles were selected initially. 43 articles excluded after reviewing abstracts and titles. 49 papers selected for which full text was managed. 06 extra papers found manually from references. 55 articles with full text were eligible for study. 24 inadequate articles excluded in final. 31 studies were finally selected for systemic review (Fig. 1 and Table 1)
Overview of included Studies
The studies included in this study were conducted between 2004 and 2021. The studies were conducted in countries like USA, New Zealand, Australia, Norway, Canada, Denmarch, Spain, Switzerland, Taiwan, Greece etc. The interventions used in studies were environmental sensor devices [24, 25, 39, 40], wearable sensors [15, 29, 37, 38]. These sensors were used in some studies for recording and monitoring health status. The robotic systems were found to record the vital signs and provide cognitive games, video callings and entertainment videos [13, 14, 16,17,18,19,20,21,22,23, 26,27,28, 30,31,32,33,34,35,36, 42, 43]. The controls used in the studies were living dog [13], standard care [14,15,16,17,18], plush toy cat [19]. The number of study participants varied from 4 to 415. The study settings included LTCs, nursing homes, dementia day care centres, living communities, residential care hospitals. Mean age of study participants in different studies varied from 67 years to 98 years. The average duration of exposure to intervention devices in studies varied from 30 minutes to 24 by 7 in different studies. The time intervals which were taken as reference point for follow up included 4 week follow up and it extended upto 2 years in some studies. The outcomes that were analysed in different studies were depression [14, 16, 20, 23,24,25,26, 30, 32, 34], quality of life [14, 17, 20, 26, 28, 30, 32, 34], agitation [16, 18,19,20, 23].
Methodological quality of included studies
The studies included were controlled trials [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29] and non controlled trials [30,31,32,33,34,35,36,37,38,39,40,41,42,43]. Among controlled trials there were randomized controlled trials [13, 27, 28], non randomized trial [14, 25], pilot RCT [15, 18, 19, 23], retrospective study [29]. Tables 1 and 2.
Reporting biases
Some studies showed low risk of bias [16, 18, 21, 24, 26,27,28,29,30, 32, 36] while some studies had high risk of bias [17, 19, 20, 22, 33, 34].
The effectiveness of randomized studies was evaluated using the RoB 2 probability of bias method. Most of the investigated studies showed substantial bias risk, which diminished trust in the stated outcomes. Although some research raised some doubt, domains examining bias from the randomization procedure and from preferential dissemination of outcomes generally indicated low probability of bias. The disclosure of missing empirical results and the suitability of outcome measures were the two main worries. High risk of bias was found in every study evaluated incorporating the RoB 2 technique for cluster randomized clinical trials. The scheduling of participant selection and enrollment in connection to the date of randomization, as well as any variations from the proposed intervention, have raised questions.
Several studies failed to record data for all variables and other problems included the presentation of incomplete outcome measures, incomplete assessment of outcomes, and selective discolosure of results. In the uncontrolled studies, there were few problems with sample selection, intervention categorisation, or variations from the planned intervention. Confidence in the presented results was, however, diminished by worries about the absence of data on possible confounding and bias resulting from outcome assessment in some trials. No research discussed how to deal with missing data.In other investigations, selective disclosure of the results seemed to be a problem.
Quantitative findings
In Tables 1 and 2 we show details about the interventions used in studies, their major both qualitative and as well as quantitative outcomes and also we showed which intervention was preferred.
Acceptability
The acceptance of various AI-enhanced therapies was evaluated through 3 controlled studies and 2 independent studies. Social robot acceptance was uneven and fluctuated according to individual robot and also their scenario of application (eg, clinical situation vs entertainment situation) [44, 45]. In two investigations, wearable sensor devices and environmental sensor devices were not well received, however one research received mixed reviews from LTC patients and carers [15]. Two of three carefully supervised studies examined social robotics. During a 12-week program, residents of nursing homes and hospitals expressed conflicting opinions of the robots like Guide and Cafero [14]. On the other hand, PARO was highly received and preferred to Guide in a dementia long term care facility [41]. An ecologically embedded mobility tracking system’s poor acceptance among inhabitants, who doubted its utility, was revealed by 3rd controlled experiment [15].
Caretakers, who valued its simplicity of use and capacity to guarantee older patients’ health, were more likely to find it acceptable. One of the two uncontrolled studies found that most participants found the social robotic MARIO (which attempted to lessen loneliness) comfortable, although there were flaws with look and availability as well as worries that persons with dementia would be cognizant of the robot’s speech [30]. Another uncontrolled study examined a variety of health examining sensor devices that the participants found to be unsatisfactory and condemned [38] people have complained that the wearables sensor devices were cumbersome, the bed sensors kept them awake, and the continual observation bothered them. Other research covered outcomes including the development of bonds and decreased load on caretaker that might help make something more acceptable [13].
Depression symptoms
The usefulness of AI-enhanced robots in treating depression symptoms was examined in 6 controlled experiments. There were no appreciable variations seen between control category and intervention categories in five investigations. In a short trial involving dementia patients, Liang and colleagues discovered that depression symptoms had decreased from start of study in both the PARO category and the control category [18]. Despite the fact that this impact was only noticeable after 6 weeks psychological distress rose in the PARO subgroup whereas they did not in the comparison category after 12 weeks, . There were no appreciable differences in depressed symptoms as observed between the intervention subgroup and control sub-groups in two investigations assessing AI based environmental sensor devices [24, 25]. After evaluating baseline observations and follow-up results from five uncontrolled studies investigating robots with smart home scenarios, depressed symptoms showed some remission.
Quality of life
Only 2 of 5 experimental studies that looked at the impact of AI-improved robots on life quality revealed a meaningful difference. Both studies evaluated PARO’s impact. Using a cross-over methodology, Moyle and colleagues exposed participants to PARO for duration of five weeks, followed by a three week rest interval. The standard of Life in patient suffering from Alzheimer’s Disease (QOL-AD) assessment was found to be moderately positively influenced by PARO [20]. In the second research, a genuine trained dog and PARO were put side by side in a skilled nursing facility [28]. Exposure happened twice a week for three months.
The Overall Quality of Life during Late-Stage Neurodegeneration (QUALID) scale ratings showed statistically significant variations, with the PARO subgroup scoring lower than the control sub-group on the measure. Three uncontrolled studies that looked at the effect of social robotic systems on standard of life came up with inconsistent findings, just one research suggesting a benefit.
Agitation
There were conflicting findings in five controlled trials that looked at the impact of pet robotics on agitation. While two trials failed to uncover any changes seen between control intervention categories, 3 studies discovered minor but significant reductions in restlessness [16, 18, 19, 21, 23].
Social outcomes
The impact of social robotics on loneliness was examined in two repeatable experiments and two uncontrolled trials. Controlled trials revealed that AIBO and PARO greatly reduce loneliness [13, 26]. In uncontrolled studies, it was demonstrated that PARO and NAO reduce loneliness [32, 34]. A total of five research evaluated social interaction. Communication between participants and nursing home staff improved significantly, according to a controlled trial employing PARO [18]. Following PARO therapies, interpersonal and interpersonal abilities significantly improved in two uncontrolled trials [35, 43]. Also, it was discovered that participants’ feelings of social support were improved by the robot MARIO [33].
Behavioural outcomes
Individuals who communicated with PARO dramatically increased their general activity involvement and were more visually, physically and verbally engaged, according to 2 controlled studies and 1 non-controlled trial [21, 27, 43]. Studies discovered a reduction in negative behavior36 and an improvement in behavioural encouragement in 1 controlled study and 3 non-controlled studies that examined behavioural situations [40]. In contrast to some other research that demonstrated a significant decline in cognitive and behavioural scores in this cohort, a controlled trial indicated that introduction to PARO was linked to a higher prevalence of wandering (a disease-related behaviour that can make people disoriented or confused) in dementia patients [35]. There were three researches on level of dependency on others. After employing the Guide robotics or Cafero robots, participants’ levels of reliance on others did not decrease much [14].
The extent to what environmental sensors devices can promote lack of dependency on others and influence daily life, however, is debatable [24, 39]. An environmentally implanted sensor system was the subject of a controlled research that revealed no significant variations in rate of hospitalisations [24]. However, a wearable sensor called CarePredict marketed by CarePredict, USA was related to reduced rates of hospitalisation.
Neuropsychiatric and cognitive outcomes
Cognitive performance was examined in five research. In 2 controlled studies, environmental sensor mechanisms have been shown to have no discernible impact on cognitive performance. 24,25] An independant trial, however, showed a notable increase in cognitive performance [37]. While some publications claim that PARO has no appreciable impact on cognition performance [28, 35] or neuropsychological symptoms18, others contend that PARO is linked to a reduction in nocturnal behavioural issues [28].
Results of physical capacity
PARO was used in two controlled experiments to measure muscle movements. While one study revealed no discernible effect [23], another discovered that it helped to decrease body movements [19]. Despite finding of no changes in balance, gait metrics and physical ability have been demonstrated to better in 2 controlled clinical trials using environmental sensing systems [24, 25]. Accidents of falls were not observed to be significantly affected by a worn sensor [29].
Psychological outcomes
Anxiety levels were discovered to be affected by social robots in different ways. Some studies have found no meaningful effect after exposure to PARO, while others have seen a slight reduction in anxiety [23]. NeCoRo had no discernible impact on anxiety [19]. Four experiments utilising PARO [20, 22, 31, 35] and one experiment using NAO demonstrated an improvement in emotional and mood states [34]. Having followed a PARO intervention, no appreciable effects in apathetic levels were seen [20]. However in a trial that used NAO, apathy considerably diminished [28].
Other health outcomes
PARO’s impact on the sleep experience was examined in two controlled studies, the findings of which were inconsistent. While one research found a significant decrease in midday sleeping and an elevation in daytime awake, another found no evidence of improved sleep habits [22]. A wearable sensor device for the environment and an uncontrolled trial showed a considerable increase in duration of sleep [37]. Results for blood pressure as well as heart rate are inconsistent. There were no discernible variations in these consequences between the subjects in the PARO management category and the control category according to a protracted controlled experiment [18]. Conversely, a non-controlled research found that PARO was connected to a brief, significant reduction in heart rate and blood pressure [42]. There was no discernible impact of PARO on pain episode [23], cortisol [18].
link