MARIO Project: Validation and Evidence of Service Robots for Older People with Dementia

INTRODUCTION

Worldwide, approximately 50 million people have dementia (5% of the world’s older population), and every year around 10 million new cases are diagnosed [1], with an estimated economic cost of US$ 2 trillion by 2030 [2], and severe social consequences (health care demand and family burden increasing, quality of life and well-being decreasing) [3].

Robotic technology can support older people living with dementia (PLWD) and their caregivers, being part of a caring and ethical home/nursing home/hospital package [4]. However, potential benefits as well as ethical concerns related to robot use for PLWD are increasing. Principally, ethical issues are related with the acceptance process, in addiction to reduction of human contacts, humanization of robots, loss of control, and infantilization [5]. The acceptance issues are mainly related to the subject attitude toward new technologies [6, 7], and the psychological factors affecting acceptance/rejection of robots are still unclear. Moreover, current literature confirms the persistence of structural limitations to successful adoption including partial lack of clinical validation and insufficient focus on patients’ needs [8–12].

In the EU funded MARIO (Managing active and healthy Aging with use of caRing servIce rObots) project [13], specific technological tools were adopted that aimed to make it easier for PLWD to accept assistance from a robot with and, in specific situations without, the presence of a human supporting the operations made by the machine.

MARIO uses the Kompaï 2 robot platform developed by Robosoft [14]. It is a robot equipped with a camera, a Kinect, and two LiDAR sensors for indoor navigation, object detection, and obstacle avoidance. A tablet PC is located on the robot torso for interaction. Mario’s controller and interface technologies support software easy plug and play development; moreover, it includes a speech recognition system to interact with natural voice during daily life. Novel IoT technologies, based on Big Data [15], are integrated to deliver behavioral skills. The MARIO robot is innovative because it integrates into a single robotic platform several capabilities supported in the literature but so far tested in isolation.

Therefore, MARIO has been designed to support and manage the following “robotic applications” (apps), also shown in Figs. 1 and 2:

OPEN IN VIEWER

Fig. 1

Overview of Apps developed during MARIO project, categorized by functionalities.

OPEN IN VIEWER

Fig. 2

MARIO robot and its user interface implemented in.

In the final stage of the project, a validation of the companion robot has been performed from August to October 2017. This paper addresses the following aims:

the impact of the apps described above, when they were delivered by MARIO;

MATERIALS AND METHODS

This multicenter study fulfilled the Declaration of Helsinki [31], guidelines for Good Clinical Practice, and the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [32]. The approval of the study for experiments using human subjects was obtained from the local Ethics Committee on human experimentation. Written informed consent for research was obtained from each PLWD, from relatives or a legal representative. PLWD were consecutively recruited in three pilot sites: 1) Residential care (National University of Ireland, Galway, Ireland); 2) Hospital (Complex Structure of Geriatrics, Casa Sollievo della Sofferenza Hospital, San Giovanni Rotondo, Italy); and 3) Community (Stockport Metropolitan Borough Council, Stockport, UK).

In total, 38 PLWD (10 in Ireland, 20 in Italy, and 8 in UK) were screened for eligibility and included in the study, according to the inclusion/exclusion criteria shown below (Table 1).

Inclusion criteria were: 1) participants with diagnosis of dementia according to the criteria of the National Institute on Aging-Alzheimer’s Association (NIAAA) [33] and the Diagnostic and Statistical Manual of Mental Disorders - Fifth Edition (DMS-5) [34]; 2) presence of cognitive impairment measured by Mini-Mental State Examination (MMSE) [35], and 3) the ability to provide an informed consent or availability of a proxy for informed consent. Exclusion criteria were: serious comorbidity, tumors and other diseases that could be causally related to cognitive impairment (ascertained blood infections, vitamin B12 deficiency, anemia, disorders of the thyroid, kidneys or liver), history of alcohol or drug abuse, head trauma, psychoactive substance use and other causes of memory impairment.

The MARIO robot was shown to all participants and the applications were demonstrated. Afterward, the researchers had time to build rapport with PLWD, described them the purpose of the study and discussed their participation (Fig. 3) They were given the opportunity to ask questions about participating in the study and then their written consent was recorded. The consents to participate were recorded at every interaction throughout the study [36].

OPEN IN VIEWER

Fig. 3

MARIO robot interacts with an elder and a clinician.

In Pre-and Post-MARIO interaction, the following parameters, explained in detail in the text below, were collected by systematic interview, clinical evaluation and review of records: demographic data, clinical and medication history and a complete multidimensional and cognitive-affective assessment.

RESULTS

DISCUSSION

In the present manuscript, we focus on the key results obtained on a small sample of PLWD across the three different pilot sites involved in the final evaluation phase of the MARIO project. The results of the OME demonstrate that engagement with MARIO varied greatly, across all three pilot sites, specifically in terms of the number of times people with dementia interacted with MARIO and in the time they spent with the robot. Participants in Ireland had more interactions than in Italy and in the UK. As Irish participants were living in a long stay care setting, they had greater access to MARIO and over a longer period. In Italy, the participants were recruited within the Geriatric unit, an acute medical ward. This has a mean hospitalization time of 7 days, which is why the number of interactions varied between 3 to 5 sessions. UK participants were required to travel to use MARIO in a specific testing venue because they lived out in the community. Because of this, researchers felt it was reasonable to request the participants’ presence on three occasions during the evaluation period. Therefore, the total time spent with MARIO in the Italian and UK sites was lower than in the Irish site.

The Irish and UK sites were found to be more accessible for MARIO interactions than the Italian site. At the Ireland site, residents were able to engage with MARIO in a variety of locations throughout the setting without being moved. A key feature of nursing homes is that in the participants are at ‘home’. Residents are free to move around the space as they engage in their daily living experiences [42]. Because interactions with MARIO could be facilitated throughout the site, the residents were less likely than participants from Italy to be moved to a secondary location. Similarly, in the UK, participants engaged with MARIO in a designated testing site, so there was no need to move to facilitate the engagements. In Italy, people are housed in a hospital room with other people; for increased privacy and to ensure a quieter environment, the participants were moved to a room specifically designated for MARIO engagements.

Participants in the residential setting were more likely to be interrupted when being invited to engage with MARIO for engaging in other activities, and to be interrupted by staff and visitors during engagements, than participants in the hospital or community setting. This was expected as other studies have highlighted nursing home residents spend the majority of their time engaging in both passive and scheduled activities [43]. People in Italy did not experience interruptions because the interactions were organized during times free from organized recreation (listening to Mass and praying) or hospital activities (doctor’s visits, laboratory and instrumental exams). Similarly, in the UK, engagements took place in a pre-booked & quiet space with no distractions.

Irish participants also refused to partake in engagements more often than the participants from other sites. This may be due to unique factors in this setting. Riedl et al. (2013) discuss the decrease in autonomy that residents often experience when moving into a long stay care setting [44]. As residents settle into routines and accept the boundaries and rules of the nursing home, they look to assert, when available, their personal choices. As consent process was put in place with every MARIO engagement, the refusal to participate in an interaction, at a given time, could be seen as the resident taking back some of their freedom to choose how they spend their time. In the hospital setting, the engagement sessions with MARIO were organized to fit within the prescribed hospital stay schedule for participant. Similarly, in the community setting, the interactions with MARIO were performed in a community hospital, and at the Alzheimer’s Society premises, where participants came into pre-arranged appointments to interact with MARIO. Since participants in Italian and the UK sites were aware that these meetings were designated for MARIO engagements, they were unlikely to decline engagements.

Requests for additional help in operating MARIO were far more prevalent in the residential care setting, where few participants needed guidance after the first engagement in both the hospital and community settings. This difference may be due to the participants in the residential care setting having a higher level of cognitive impairment, with two residents with moderate cognitive impairment and another two residents with severe cognitive impairment. In the hospital and community setting, people with mild cognitive impairment were recruited.

The Irish residents showed high levels of enthusiasm and engagement across the interactions and the UK participants showed mostly positive levels of attention and engagement. Conversely, participants in the hospital setting showed a lower level of enthusiasm and attentiveness towards MARIO. This may be because in the hospital and community settings MARIO’s speech recognition problems (in particular the software’s lack of understanding of local language/dialects) became frustrating for participants and this was a demotivating factor that negatively affected the participant’s engagements with MARIO. In the nursing home setting, the researchers turned off the voice activation. This was done because the noise levels in the surrounding environment made it difficult for MARIO to hear the low speech volume of the residents. Instead, residents operated MARIO through the touchscreen and or using a stylus, in case they found this easier than using their fingers. Modifying activities so they are tailored to the abilities of the individual has been found to increase engagement in people with dementia [45]. This may explain why Irish participants exhibited higher levels of attention and a more positive attitude than participants from Italy and the UK.

In the hospital and community settings, most participants were found as having no, or only slight, cognitive and dexterous difficulties when operating MARIO. The residential care setting participants varied on the levels of difficulty they experienced. But they still reported mostly slight to moderate issues. These differences may be explained by the cognitive impairment disparities of the participants recruited in the three pilot sites. This is congruent with the results from other studies which found cognitive profile to be an influential factor in technology usability, and the ability of participants with mild cognitive impairment to complete tasks, due to a decline in mental capacities such as attention and perception [46, 47].

The Irish participants most often interacted with MARIO by dancing with it or singing along with the music from the MyMusic app and reminiscing with staff and relatives via the MyMemories app. While those in the hospital and community settings were more likely to spend their time directly operating and manipulating the applications, the most used applications being similar to favorites from Ireland, including MyMusic, MyMemories, and MyGames. Along with the applications favored by the other sites, Italian people also utilized the CGA app a lot (average time of 25.30 min a day), and MyCalendar app (with an average time of 24.15 min a day). Across all sites, MARIO was found to facilitate conversations and social engagement with staff and relatives and provided participants the opportunity to talk about their own life.

The results reveal that there was a significant improvement in the resilience level of participants between the pre-and post-interaction measures (p = 0.020). The increase in the resilience of participants following MARIO engagements was proved through noticeable improvements in mood and focus. The data also found that participants reported an increase in their QoL scores. This suggests that in spite of their cognitive decline, participants kept and/or improved their autonomy in activities of daily living, their social interactions with others, and they experienced more positive emotions. These results are comparable to those obtained in recent studies: a study using the AIBO robot dog found improved health-related QoL in PLWD [48]; and a pilot RCT study using the Paro seal robot revealed a moderate to large clinical influence on QoL in 18 PLWD [49]. It is also not surprising that these participants also reported a decrease in their depressive symptoms. Thanks to the findings in the current study in terms of improvement in the QoL, MARIO could potentially enable participants to slow down the progression of their dementia, enabling them to maintain their level of psychological and physical functioning and their capacity to independently manage their activities of daily living [50]. Moreover, maintaining relations with others and having control of life are aspects that influence QoL in dementia [43].

Given the total sample size was small (n = 38), and the study had a relatively short length, researchers did not expect significance levels to reach the customary p < 0.05 [51]. This suggests that any statistically significant results must be interpreted with caution. Nevertheless, these findings indicate that MARIO may be a useful tool in mitigating depression and loneliness while enhancing social connectedness, resilience, and overall quality of life for people with dementia. Further research using a larger sample, is required to investigate this potential.

Anyway, the MARIO platform aims at becoming a pilot experience to design significantly personalized robotic applications while reducing development costs and the response time to address existing and emerging people needs. From a clinical point of view, it is an enabler to collect data that can improve medical treatments and care personalization. Specifically, data coming from conversations (i.e., patients’ answers) are extracted by MARIO through a Speech-to-Text module. Thus, it is possible to convey the information into an Ontology network. Although there is no standard ontology that can be used as a base for robot semantics in this field, the MARIO project has developed a specific Mario Ontology Network (MON), evolved by integrating ontologies emerging from interactions with humans, on board and external sensors [52]. The use of semantic data analytics and personal interaction has tailored the applications to better connect older persons to their care providers, community, own social circle and also to their personal interests. In particular, the integration of robot semantics with existing structured and unstructured data leveraged on innovative data integration practices (e.g., W3C semantic web, ontologies, etc.). The knowledge management is Entity-centric, that is, each entity and its relations have a public identity that provides a first grounding to the knowledge used by robots. The networked ontologies are used for organizing the stored data and support internal processes.

Moreover, the use of such diversified apps gives the clinical team the opportunity to use the MON with personal information to connect various aspects of the patient’s status and recovery activities over time.

In a future perspective, further functions could be implemented and interesting reports could be brought out by MARIO apps in order to obtain increasing amounts of data in user behaviors for future new Big Data studies in the field of Personalized Healthcare.