Abstract
Keywords
INTRODUCTION
Prevalence of dementia and the necessity of assistive technology
Alzheimer’s disease (AD) is a progressive neurodegenerative condition that affects memory, cognitive function, and daily activities and, thus, it is one of the most important contributors to dependence, disability, and cost to society (818 billion US dollars in 2015) worldwide [1, 2]. Nevertheless, its prevalence is increasing as the number of people age 65 and older with AD may nearly triple by 2050, from 46.8 million to 131 million people around the world, the majority of which, living in an institution [1]. The disease creates new economic, social, and health care demands and, as it gets worse, more attention is being given to the role of assistive technologies [3].
Assistive technologies could fill an important needby improving clinicians’ diagnosis and decision-making in order to meet an individual’s needs and could be used as an objective measure of cognitive status and disease progression rather than neuropsychological methods. Assistive Living and Monitoring, a term used to describe a smart system, is “a home equipped with technology that allows monitoring of its residents, provides objective and continuous observations and promotes independence and the maintenance of good health” [4]. Additionally, assistive technology is expected to play an important role in supporting clinicians and improving quality of life as well as cognitive and physical state of people with dementia (PwD). Drawbacks of current health services are that they often aim to evaluate single needs (e.g., pharmacological treatment) and detect single problems via brief interviews instead of assessing, intervening, and evaluating multiple problem areas in various stages of the disease, leading to generic interventions not regularly evaluated.
Home remote monitoring of patients is a promising “participant-centered” management approach that provides specific and reliable data enabling clinicians to provide adaptive and personalized interventions in order to improve patients’ physical and cognitive condition by monitoring their daily function. An effective clinical intervention through smart systems requires participant’s self-monitoring (e.g., daily sleep observation) and a clinician who is keen on empowering him/her. On the other hand, clinical psychosocial interventions, which can be delivered to people with dementia at home, address symptoms and needs common in dementia spectrum. “Cognitive interventions may involve many approaches aimed at restoration of function, implementation of compensatory strategies and environmental modification, and these can be integrated with approaches directed at dealing with the emotional responses to impairment and other psychosocial difficulties to provide an holistic neuropsychological rehabilitation framework” [5]. In general, clinical interventions involve (i) information about the cognitive state, strategies to “forestall its progress” and effective treatments through innovative and supportive systems; (ii) training in managing the disease, including how to self-monitor symptoms (e.g., sleep problems, medication), setting aims and goals, and developing confidence; (iii) behavior and physical modification programs such as physical exercises; and (iv) counseling, advice, and support to help participants cope emotionally with their daily difficulties. Also, communication between clinician and caregivers as part of intervention may also have positive effect on managing a participant’s health, because it provides more accurate and updated information and helps the clinicians to make better decisions regarding the clinical interventions.
Therefore, assistive technologies are able to address the needs of PwD. Smart systems can improve the quality and variety of information monitored from specific measures of physiological signs and behavioral patterns and translate them into accurate predictors of health condition and disease progression.
Why information and communication technology-assisted living, monitoring, and interventions?
Assistive Living in general has been designed to monitor a home, based on environmental and wearable sensors, smart devices, and appliances (tablets, etc.) for various reasons such as: safety, activity monitoring, providing prompts and reminders, fall detection, physiological monitoring, and vital signs (heart rate, calories, steps, etc.). In general, a sensor-based system manages and assesses daily activities and provides to the clinician and/or the caregiver remote monitoring capabilities. Several efforts aim to highlight the importance of using assistive living in people with dementia.
A great deal of existing assistive information and communication technology approaches for the care of elders in general and people with AD in particular, utilize a single or a few devices and focus on specific problematic areas and domains such as wandering, falls, sleep, or daily tasks. Specifically, some studies [6, 7] focus on “monitoring and surveillance”, aiding participants with AD not to get lost by using GPS devices and provide an enhanced feeling of safety for them and their relatives [8–11]. Other types of assistive technology which prevent falls [12, 13] have been found to be beneficial because it addresses specific needs of the caregivers [9–11]. Additionally, there are efforts that share physical activity levels with the clinicians or the family members, by monitoring with sensors the motor and cognitive activity of elders [14], while others assess residents’ performance of daily activities and behavioral aspects on everyday tasks [15–17]. Other systems focus on the medication management by using instrumented pill containers [18], observe and assist hand washing in people with moderate dementia [19], or collect data from sensors embedded in the kitchen while participant performs habitual specific tasks, such as making coffee [20]. There are few existing studies, NOCTURNAL and Rosetta, which have focused on sleep and have conducted a relevant longitude study, providing both nighttime activity and home monitoring observations [21, 22].
More complex solutions involve smart home deployments of environmental sensors to observe and assess elder and disabled people’s activities [16, 23]. Such systems have been used also in the ‘Welfare Techno-Houses’ in order to monitor the residents’ physical activity and vital signs with wearable and presence in the bathroom with door sensors and “fully automated biomedical devices” [24], while other systems provide security by using actuators to control doors, windows, and curtains [25]. However, none of the above solutions monitors and records the most important physical activity of the participants, which is the night sleep activity. Additionally, more severe levels of impairment are targeted with sensors to recognize and assess not only physical activity but also weight and vital signs by providing alarms in case of emergencies [26–28]. Other systems help people with cognitive deficits to remember and carry out their daily activities (e.g., instructions how to make coffee, make a phone call, etc.) and ensure safety but they do not affect their quality of life or their everyday functionality [29]. Moreover, sensors have been attached to electrical appliances (a cooker, refrigerator, TV, etc.) to detect events caused by disease or possible accidents [30]. For example, ‘Aware Home’ which aims to assist declining memory in elders, recognizes potential crises and identify behavioral trends [31]. Finally, there is a number of assistive technology approaches which perform clinical assessment of functionality and cognition of participants with technology [32] or improve impaired physical functions [33].
However, one common limitation across the existing approaches is their focus on a single domain among physical activity, sleep, wandering, medication, and daily activities. Moreover, none of the existing assistive technology approaches has applied adaptive interventions in order to support participants, improve their cognitive function, and tackle behavioral problems. Although some approaches provide a user-friendly interface for clinicians and caregivers, very few can provide insight to all aspects and domains of interest along with their relation to one another. There have been reported systems with motion sensors which have tested mobility patterns related to cognitive ability [34], embedded sensing systems which recognized changes that are associated with cognitive decline [35], and others which have used environmental sensors, including motion detectors and magnetic door sensors, to gather information about complex activities such as cooking, cleaning, and eating [13, 37].
Our system systematically addresses all above issues and is able to a) combine information coming from several types of sensors which are used for recognizing moving intensity, sleep patterns, and presence in specific home areas and b) perform reasoning on this combined information in order to monitor and assess high-level daily activities, which cannot be identified only by a single sensor. It provides holistic monitoring of all areas of interest, specific visualizations to the clinicians/caregivers/participants, and technology-aided personalized interventions in order to improve participants’ condition while maintaining human clinician and participant contact through clinical interventions.
In detail, on one hand, the system collects multi-sensor observations and records of daily activity performance of participants, but on the other hand, utilizes deep-insight monitoring for effective, tailored weekly clinical visits, interventions, and progress updates. Moreover, in contrast to other approaches, our system is based on multi-sensor fusion and interpretation through intelligent reasoning mechanisms and semantic knowledge structures, providing clinicians with a comprehensive image of the person’s ability in daily activities and overall behavior expressed in daily activities modeling through person-tailored user interfaces (Fig. 1). More specifically, the proposed system identifies potentially problematic areas across several domains (sleep, physical activity, activities of daily living (ADLs), mood, and social interaction) using individualized problem-detection parameters, and examines these patterns to identify cognitive improvement and decrease of behavioral problems. This helps the clinician, who interacts face-to-face with the participant or via messages through the user interface, to have a deeper knowledge about everyday activity performance. Also, specific interventions were designed to address possible problematic areas and focus directly on real-life, as monitored by the system, everyday situations, by using “evidence-based” psychological methods. The direct observations of the participant and the relevant adaptive interventions have been found from our system to be the most ecologically valid approach from the assistive technology spectrum. Moreover, despite the fact that dementia currently cannot be pharmaceutically treated, the interventions can “truly live up” to participants’ potential cognitive impairment management and slightly cognitive improvement through clinical interventions seems more effective with the help of both smart technology and the clinicians.
The deployment of the system in four different homes with full set of sensors in order to monitor ADLs was the main purpose of this research. Our main target was to have full data collection from all the homes where participants lived on their own, to introduce clinical non-pharmacological interventions, to enhance participants’ quality of life and improve their cognitive functions and functionality based on clinical assessment after a 4-months trial, and to measure the acceptability of the system (installation, sensors) both from the participants and their caregivers. Moreover, we examined if our system is able to detect changes in participants that were caused by the clinical interventions. We also introduced the end-user and caregiver interface both to the participants and the caregivers in order to be informed on a daily basis regarding their physical activity, sleep quality, and instrumental ADLs (IADLs) and to receive messages/reminders from the clinician.
In general, our research focused on high priority requirements for the home environment including the interface for the clinician, the participants, and their caregivers, the fusion of data from various embedded sensors, the visualization of these data in a way that was simple and easy for clinicians to use, and the application of personalized adaptive interventions. We selected specific sensors based on participants’ needs in order to monitor the most important, problematic, and needs-to-be-addressed areas. Additionally, we explored the utility of wearable sensors to recognize changes in moving intensity for specific activities, such as washing dishes, meal preparation, night-time sleep, taking medicine, and using the phone. These activities are characterized by interacting with unique objects and/or by the presence of the participant in the area of monitoring and have been explored by using tags, sleep sensor, presence and motion sensors for tracking these objects, and using the data for activity recognition. Moreover, during the first stage of the intervention period, the clinician identified the most prominent everyday difficulties experienced by the participant, and together with the participant set therapeutic goals relating to these difficulties. There is growing evidence that non-pharmaceutical interventions seem to be the only possible solution, until recently, for amelioration of cognitive deterioration. Thus, direct observation of the patient and adaptive interventions, in order to help the patient to become more cognitively active and to determine everyday functional status, has been found from our system to be the most ecologically valid approach enabled by assistive technology. We have selected specific adaptive interventions based on patients’ characteristics, their cognitive limitations, and their progress during the trial period. Finally, regarding the participants, we strongly believe that for clinical researchers who focus on the field of dementia, it is extremely important to take into account different perspectives of the dementia spectrum. Even if participants have same diagnosis and similar cognitive and physical limitations, they act in completely different ways in their daily lives. Thus, the future of addressing and forestalling cognitive impairment is to focus on more personalized solutions, which will answer more dilemmas regarding the heterogeneity of the disease.
Therefore, the paper has three main objectives: (i) to provide a monitoring (AAL) system for continuous and objective monitoring of problematic daily living activity areas, which consists of sensors installed at home, data processing, and an user interface for visualizing information to clinicians/participants/caregivers; (ii) to design and update accordingly appropriate and personalized interventions based on system feedback and clinical observations to improve and restore cognitive function and health-related quality of life; and (iii) to measure and evaluate the effectiveness of the interventions in four cases for at least three months in order to identify if the applied interventions affect positively the participants’ modalities (e.g., sleep, moving intensity, ADLs).
METHODS
Case studies
The recruited participants were selected based on the NINCDS-ADRDA [38] for dementia and Petersen criteria [39] for mild cognitive impairment (MCI) and deemed to be fit to take part in the study by a neurologist specialized in the field of dementia and a psychologist. All the participants were recruited from the Alzheimer Day Care center in Thessaloniki, Greece, and ethics approval was given by the Alzheimer Association Scientific and Ethics Committee. We focused on stages 3–5 of the disease according to Global Deterioration Scale. We have recruited 4 participants (2 amnestic MCI (aMCI) and 2 PwD). The first three were involved in a4-month period and the forth in a 3-month period trial (Table 1).
A comprehensive neuropsychological assessment showed that his executive functions were
below the mean for people of his age and education level (12 years), and his memory
storage was impaired. Furthermore, the participant had difficulties in cognitive speed
and processing, and the main difficulties observed in the participant were found to
include item selection and passivity in initiating actions
autonomously.
memory problems and difficulties in finance and medication management. During clinical
assessment, the participant looked anxious and nervous, and slow speech processing was
also detected. She met the criteria for mild depression. She made complaints of
cognitive deficits and was facing difficulties in activities of daily living such as
cooking and housework.
Even though we did not expect improvement in cognition during the observation period
due to the progressive nature of the disease, we wanted to support the caregivers’
awareness about the participant’s medication and alertness. Her caregiver mentioned the
following issues: 1) memory deficits; 2) difficulties in activities of daily living
(e.g., cooking, housework, medication); 3) limited social contact; 4) the need to
monitor her health, especially when they were out of town; and 5) reduced competency in
managing medication and finance.
The system
The proposed system is a comprehensive assistive technology home care solution to support care and prolong independent living of people with dementia. The system utilizes unobtrusive, online sensors, mobile devices for feedback and intelligent analysis, in an Ambient Assisted Living (AAL) context. While in-depth system specifications are outside the scope of this paper (which concerns clinical enablement and results), its fundamental properties and capabilities are presented below. The detailed architecture, as developed in the framework of the Dem@Care FP7 1 project can be found [40, 41].
The overall system architecture, as shown in Fig. 2, aims to unify a wide variety of sensors and hardware to enable multi-domain monitoring, adaptable to each individual’s needs. One of its main features is that it unifies and repurposes proprietary, affordable sensors (hardware) in a medical context. In each installation, a selection of multiple sensors covers almost every possible modality, streaming measurements to an in-home gateway. Specifically, a wristband 2 , worn 24/7 and charged during weekly clinical visits, monitors Moving Intensity per minute. A standard IP camera 3 streams images for analysis (not stored). An under-mat sleep sensor 4 measures sleep quality. Two unobtrusive sensor networks are deployed: a set of smart plugs 5 (power consumption sensors) monitors utility usage of appliances and a set of smart tags 6 (accelerometers) detects object usage through accelerometer motion and human presence through infrared motion detection. An overview of all modalities is given in Fig. 2.
The system ensures harmonization of data provided by the underlying modules on a semantic level. A module for each sensor ensures retrieving its data through manufacturer-specific and platform-dependent APIs. Consequently, heterogeneous sensor data is streamed, processed, and stored in the Knowledge Base of the system, in unanimous format, using our common vocabulary for their semantics, i.e., an ontology, which ensures interoperability within and outside the borders of our system. These underlying processing techniques range from simple retrieval of sensor measurements to complex human activity recognition methods based on image data, as presented in [42]. Consequently, all unanimously described data in the knowledge base are iteratively fused, using semantic interpretation and intelligent decision making techniques, to derive higher-level complex activities and detect clinical problems. For instance, the framework implements a rule-based higher-level fusion method that detects the activity of making tea by fusing information about the location of the person detected by presence sensors (e.g., in the kitchen), the objects the person interacts with using power consumption and motion sensors attached to devices (e.g., the kettle), and primitive events detected by ambient cameras (e.g., drinking). This is depicted in Fig. 3, where all detected events are shown on a timeline. Utility usage (shown in green) is combined with object usage, presence (in blue), and event recognition (not shown) to generate high-level events (in purple). For instance, on Fig. 3, having the TV on and moving the remote constitutes a watching TV activity. In this way, the clinician is provided with intelligent detection of complex events instead of carving them out himself. He can also choose to view an overview of events (filtering out low-level ones) or monitor all events in detail. Likewise, several rules have been implemented for problem detection, e.g., to derive a “short sleep duration” problem if total sleep for a night is less than eight hours. Such rules in both cases are personalized for each participant and iteratively refined as the pilot progresses.
According to the needs and peculiarities of each individual, a deployment may be adapted on-the-fly either in terms of modules, i.e., kinds of sensors, or the number of sensors. In the first case, removing sensors translates to simply not introducing them in the intervention as, for example, an individual may not present interest in sleep or activity related interventions and, hence, not wear the wristband or use the sleep sensor, respectively. Adding modules is still possible through the system’s service-oriented infrastructure and semantic interoperability, although not on-the-fly (new plug-in module needs to be developed). This is essential to serve different households and intervention interests as some individuals may use more devices and objects for one activity than others, e.g., one might use microwaves, a cooker, and a boiler for cooking or have two bathrooms. Thus, in some cases, more plugs and more tags are needed. This feature, scaling up the number of sensors (of the kinds currently supported) is supported by the system on-the-fly (at deploy-time), as the chosen wireless sensor networks of smart plugs and smart tags, themselves already support it. In general, the system supports infinite cameras, one wristband, one sleep sensor, and up to around forty tags and plugs per deployment. While these limitations may require some effort to overcome, we have found them to be more than sufficient for all four proof-of-concept deployments presented in this paper.
Finally, all information is displayed in tailored user applications for each of the involved entities, namely the Clinician and the End-user & Caregiver interfaces. The former application presents the totality of information, statistical means and deviations, in configurable views, where a clinician can specify resolution, date range, filters per activity type or sensor and perform comparisons. Utterly, he/she is able to observe patterns and trends either with respect to monitored qualities or problems detected by the system. The latter application, installed at a touchscreen tablet, allows end-users (if they are willing and capable to do so) and their caregivers to access a reduced view of monitored qualities, such as sleep, physical activity, and some usage of utilities, e.g., TV. The visualization hides problems and gives a positive outlook of what has been achieved (rather than not reached), endorsing confidence. Fig. 3 gives an overview of the visualized information. Both applications are extensively presented in the rest of the paper as most outcomes are presented on figures directly captured from the clinician and End-user & Caregiver interfaces.
Regarding the cost, each full system home deployment costs about 1700€ (with 100€ variation depending on the setup). The system is using affordable, retail sensors but, still, it was built for research purposes and the cost is not optimized for multiple deployments. In practical installations, less sensors might be used per participants depending on their needs. It is expected that a basic version still providing valuable insights can be reduced to 700€.
Experimental conditions
Cognitive assessment and clinical interventions
In the first phase of the intervention, the psychologist identified the most prominent everyday difficulties experienced by the participants, and together with them and their caregivers set therapeutic goals relating to these difficulties. In line with the main areas of interest, this process was structured around the areas of mood, sleep and physical activity, cognitive problems, social interaction, and ADLs. Psychometric measures relating to each area were carried out with the participant and their caregivers. The pre-test neuropsychological assessment was conducted between 1 and 2 weeks prior to the interventions, while the post-test was conducted between 1 and 2 weeks following the end of the observational period and sensor installation (Table 1). Each time, clinical and neuropsychological assessments were taken in one session that lasted approximately 60 minutes (applied standardized tests are shown in Table 2). In addition to neuropsychological assessment based on face-to-face semi-structured interviews with a psychologist, the participant answered questions about his/her clinical history and medication treatment, questions about his/her childhood and his/her family, and described the relationship with his/her (wife/husband) and children, as well as what is the major problem he/she was facing. In addition, more specific questions to the participant were also conducted (e.g., “Tell me what would help you feel more safe at home during the night-time and daytime?”, “How is your sleep at night?”, “When you open your eyes in the morning how do you feel and what is the first thing comes up to your mind?”, “Do you feel satisfied with your life routine?” etc.). After clinical and neuropsychological assessment, the consensus of professional researchers (authors 2–6) installed the home monitoring system according to personalized preferences and needs of participants. Afterwards, specific adapted interventions were introduced based on the clinician’s observations through the system, and weekly scheduled meetings were also set with participant and clinician in order to adjust the interventions according to the participants’ needs. During the meetings, the clinician prepared specific topics based on the participant’s needs to be covered. The psychologist revisited participants on a regular basis and explored how they were interacting with the system or if faced any technical problem. Once therapeutic goals were set, the clinician designed and implemented strategies to address them. Each participant followed various non-pharmacological interventions in order to enhance their cognition, physical health, and behavior. The areas of emphasis were derived from a) a review of the problems typically encountered by the person with cognitive impairment; b) input from multidisciplinary dementia care experts, c) the system’s data analysis that was available to the clinician; and d) the participant’s preferences and needs after guided advances from clinician.
The main aim of the intervention was to improve different functions and to activate and motivate participants to enhance their cognitive health behaviors by remaining cognitively active and compensating for deficits. The intervention protocol was structured in main phases: The psychologist visited the participants’ home at a time that was convenient to them (Week 1-Visit 1), explained the project’s regime to them and their caregiver, explained the daily diary in which the participants had to keep notes, especially during changes in the schedule. During the second phase (Week 2-Week 15, 2 visits per week), the psychologist introduced the package of the adaptive interventions for each participant based on cognitive and behavioral limitations as found from both neuropsychological assessment and the system’s visualizations. In addition to psychological interventions, participants followed an exercise program twice a week for two hours. The third phase (Week 16- Final visit) included the evaluation of the project. The mean length of home visits was 40 min(Table 3).
In general, we followed nonpharmacological interventions which have been suggested in literature as possible tools to support people with dementia to manage complex behaviors and improve their quality of life [43, 44] such as: cognitive rehabilitation which focuses on enabling people to engage in everyday activities and identifying specific strategies to deal with the difficulties resulting from changes in memory such as problem-solving, communication techniques, and simplification approaches (visual cueing, placing objects in sight or out-of-sight). Engagement of patients in daily exercise and use of memory boards for cueing and orientation. Additionally, approaches such as cognitive behavioral therapy to address anxiety in Alzheimer’s disease (CBT-AD), group psychotherapy, and relaxation exercises were employed which integrate empirically supported interventions for late-life anxiety and depression with strategies that facilitate comprehension, encoding, and retrieval (e.g., memory cueing, prompts).
The applied interventions included Reminiscence therapy once a week [45], inducing a vocal recall of past activities, events, and experiences in the life of a person by using tangible prompts in order to improve general cognitive function and more specifically to ameliorate retrospective memory problems, depression, and anxiety symptoms. Some memory triggers such as photographs, foods, music, household and other familiar items from the past were also used as described in our previous work [46]. Psychotherapy sessions were also applied once a week, which tailored specifically to the treatment of depression as a non-pharmacological intervention and which was our second intervention approach, based on CBT-AD principles [47] in combination with a systematic relaxation technique in order to eliminate the anxiety symptoms and depression. Additionally, for our first and third participants, Group psychotherapy, which has been demonstrated to be an effective treatment for numerous psychological difficulties and the management of behavioral disorders, was also introduced once a week in Alzheimer’s Daily Center. Treatment components included awareness training, breathing skills, coping self-statements, behavioral activation, and sleep skills. Also, due to the fact that there is strong evidence that a physically active life is beneficial and indicates positive effects on mental health outcomes in adults, the clinician expected that Physical Exercise is able to reduce depression and anxiety in participants with MCI and dementia and improve both their cognitive and physical health [48]. Thus, as an additional type of intervention, a two-hour program once a week with physical exercises was introduced to all the participants as a strategy for increasing pleasure and as a distraction from an anxious mood. The intervention exercise program delivered an individually tailored regime of walking designed to become progressively intense and last between 20–30 minutes following a similar protocol previously described [49]. These activities were assigned as homework. In addition to the specific interventions just described, another target of therapy for our first and third participants were Dance lessons for 40 minutes twice a week as followed in [50, 51].
Moreover, memory exercises were added to a participant’s weekly schedule once a week based on cognitive rehabilitation strategies as described in [43, 52]. An exercise, which included movies from the old Greek cinema, was followed by specific questions, also introduced in order to enhance participants’ ability to remember facts, faces, and information from a movie. Also other memory exercises including semantic production of words and construction or competence of sentences were introduced to the second and fourth participants. At a specific visit from the clinician, a white memory board, which was divided into four sections: “likes”, “dislikes”, “targets”, and “things-to-do”, was introduced to all participants. The participants were asked to fill in the sections during the week and note activities that they would like to do and activities they had to do. On the next visit, the clinician and the participants discussed and adjusted or changed the written notes on the white board and next targets were set accordingly. Finally, reality orientation therapy, which is a strategy for people with cognitive impairment with memory problems and “time–space disorientation” [53] was applied to our second and fourth participants, and the main goal was to improve their space and time orientation by using prompts and other memory aids [54, 55]. Specific prompts and aids were used as a medication reminder for the participant. We used cueing strategies that are helpful for the participants to initiate an action as has been proposed in [43] for other activities (e.g., effective communication strategies). Meal preparation and kitchen cleaning was one of the main problems for the fourth participant and one of the main caregiver’s concerns was safety (e.g., leaving the oven turned on). In order to address this issue, we introduced a list with specific tasks that the participant has to do every time she started cooking and she had to mark ‘Yes’ or ‘No’ in every step (Table 4). The construction of this intervention was based on a diary of a recent intervention study which focused on daily exercise of PwD [56] and similar strategies by using cues and steps for PwD as demonstrated in [43].
Clinical observations and correlation between variables in clinician interface
The Clinician interface provided a categorization of the data (e.g., activity, ADLs, sleep, etc.). Below a detailed description of the measurements and information that were made available through the Clinician interface is given. This information was used by the clinician to monitor a participant’s sleep and daily activities in order to adjust and apply specific interventions.
Sleep measurements
In the interface, the clinician can see the sleep in four separate sections. In the Summary One-Day, only values from one night appear with specific sleep stages and duration (total time in Bed Awake, total time deep sleep, total time shallow sleep, total time asleep, sleep latency, and number of interruptions) as can be seen in Fig. 4. In the Summary per Day section, the clinician is able to select more than one day and observe sleep stages and other ADLs, such as bathroom visits during the night. For instance, the clinician could detect through the system the activation of motion sensors in the bathroom while the sleep sensor recorded that the participant was awake at that specific time as shown in Fig. 5. Also, in the Dashboard section, the clinician can set specific thresholds in order to identify sleeping problems during the night (e.g., the participant has more than 4 interruptions during the night), and in the Summary per Day section, these thresholds can be presented as shown in Fig. 6. These thresholds are used by the decision-making module that semantically enriches and interprets the detected activities for the derivation of abnormal behavior and clinically relevant situations. This is achieved by coupling the recognized activities with a context-aware decision making layer that encapsulates clinical and person-tailored parameters and thresholds for monitoring and assessing behavioral aspects. Finally, in the Comparison per Day or per Week section, the clinician can display and control the whole observational period by selecting specific days or weeks, in order to see how a sleep pattern affects another one (e.g., Number of Interruptions and Sleep Latency) or how physical activity or daily living activities (e.g., Moving Intensity, house work) affect a sleep pattern (e.g., Total time Asleep), as shown in Fig. 7.
Measurement of activity and ADLs and comparisons with other measures
Difficulties in performing ADLs at home may indicate the need for personal assistance or relocation to residential care settings. Similarly, the main difficulties observed in the demented participants were found to include item selection and passivity in initiating actions autonomously. Our participants had very decreased moving intensity in the initial period and they did not get involved in housework such as cooking or cleaning. They preferred to stay at home and watch TV for many hours. Therefore, it was decided that visualization of daily moving intensity and monitoring of daily activities could provide clinicians with relevant and useful information regarding the progress of the participants in time. While viewing daily monitoring events in the Summary One-Day section despite sleep patterns, the clinician is able to check everyday activity and moving intensity for one day (Fig. 8). Also, in the Comparison per Day or per Week section, correlations between different sleep patterns (e.g., number of interruptions and total time asleep) and comparisons between electrical appliances (smart plugs and tag sensors) could also provide clinician with useful information regarding daily activities (Fig. 9a) and if a prolonged activity (e.g., TV use) negatively affects another activity (e.g., sleep) (Fig. 9b).
End-user and caregiver interface
The clinician provided to the participant (first and third participants) or their caregivers (second and fourth participants) a mobile tablet device and introduced them to the participant and caregiver interface. There were iterative learning sessions in which the clinician presented to the participant the operation and the information that the interface was able to provide.
The first and third participants were akin to technology and have enjoyed reviewing their daily measurements. Naturally, a limited view of the measurements was displayed to them in order to avoid overwhelming them or even causing stress in case of small irregularities. Educational material such as recipes, nightly routines, etc., were also included. Reminders were sent regarding the participantspecific activities (e.g., medication) and finally, messages and reminders are exchanged between the participants and the clinicians to enhance their daily routine without missing activities. Figure 10a shows a combined view of the message inbox (top) and sensor readings (bottom). Also problems of the Number of Interruptions appear in the red line above the chart, which inform the participants as to how many interruptions they had (“You had 7 sleep interruptions on 30/10/2015”). Also, the user interface provided information regarding the daily steps and total burned calories (Fig. 10b), the sleep duration and interruptions, the device usages (Fig. 10c), and medication (Fig. 10d). Overall, the application has not only helped the participants and their caregivers to feel more confident and secure with the system, but also encouraged the social interactions between them and the clinicians.
RESULTS
Neuropsychological results
Statistical analysis was conducted using SPSS v21.0 (IBM Corp., Armork, NY) statistical software. The Shapiro-Wilk test was used to assess the normality assumption for continuous variables. A pair sample t-test with the pre and post neuropsychological assessment for all the participants revealed that there was improvement in the majority of scales but statistical significant better performance was detected in Rivermead Behavioral Memory Test (RBMT)-story direct recall in final assessment (M = 10.88, SD = 4.66) than in the initial (M = 9.75, SD = 4.86);t (3) = –3.58, p = 0.03, in MMSE in final assessment (M = 28.00, SD = 2.3) than in the initial (M = 25.25, SD = 3.86); t (3) = –3.22, p = 0.04, in Hamilton Depression Rating Scale final assessment (M = 4.25, SD = 2.21) than in the initial (M = 11.00, SD = 4.08); t (3) = 5.4, p = 0.01, in MoCA final assessment (M = 24.5, SD = 4.51) than in the initial (M = 19.25, SD = 4.99); t (3) = –8.34, p = 0.004, in Rey Auditory Verbal Learning Test total score final assessment (M = 42.00, SD = 18.35) than in the initial (M = 34.75, SD = 15.35); t (3) = –3.46, p = 0.04 (Fig. 11).
Sleep
In order to investigate our hypothesis that the proposed system and the adaptive clinical interventions can have positive effects on participants’ physical and cognitive function, paired-sample t-test analysis was used to find any improvement and changes between the initial and final observational periods of a participant’s sleep, physical activity, and ADLs. The level of significance we set was p = 0.05. In our first participant, there was a significant increase in the duration of total time asleep after the observational period (M = 8.14, SD = 1.32) than in the initial (M = 7.24, SD = 2.40); t (81) = –3.44, p = 0.001, decrease of sleep latency after the observational period (M = 3.33 min, SD = 4.05) than in the initial (M = 6.83 min, SD = 10.7); t (80) = 2.7, p = 0.009, increase of deep sleep duration after the observational period (M = 1.94, SD = 0.96) than in the initial (M = 1.61, SD = 0.75); t (83) = 2.304, p = 0.02, and decrease of shallow sleep duration after the observational period (M = 3.88, SD = 1.33) than in the initial (M = 4.36, SD = 1.34); t (74) = –2.54, p = 0.01. Although, the number of interruptions decreased after the observational period, there was not a statistical significant difference.
In the second home, statistical significant differences were revealed in an increase of total time deep sleep duration after the observational period (M = 1.22, SD = 0.59) than in the initial (M = 0.90, SD = 0.40); t (59) = –3.74, p = 0.000, increase of total time asleep after the observational period (M = 7.36, SD = 1.39) than in the initial (M = 6.40, SD = 1.23);t (59) = –4.5, p = 0.000, and decrease of shallow sleep duration after the observational period (M = 3.04, SD = 1.07) than in the initial (M = 2.31, SD = 1.16);t (57) = –2.4, p = 0.02. Although the duration of REM sleep activity is longer in the final observational period and the number of interruptions and sleep latency duration are also decreased, there was not statistical significant difference between initial and final period.
Regarding the third home, there was a significant decrease of number of interruptions after the observational period (M = 3.8, SD = 2.23) than in the initial (M = 5.8, SD = 3.23); t (45) = 3.4, p = 0.001, decrease of sleep latency after the observational period (M = 8.4 min, SD = 13.15) than in the initial (M = 8.8, SD = 10.06); t (49) = 0.15, p = 0.88, and increase in deep sleep duration after the observationalperiod (M = 2.13, SD = 0.68) than in the initial (M = 1.83, SD = 0.64); t (44) = –2.12, p = 0.04. Despite the fact that no statistical significance regarding total time asleep and shallow sleep duration after the observational period was observed, there was an improvement in the final period in bothvariables.
In the fourth participant, there was a significant decrease in the number of interruptions after the observational period (M = 2.25, SD = 2.1) than in the initial (M = 3.7, SD = 3.6); t (34) = 2.29, p = 0.02, decrease of sleep latency after the observational period (M = 5.1 min, SD = 8.22) than in the initial (M = 10.1, SD = 10.8); t (33) = –2.16, p = 0.03, but no statistical significant differences were found regarding total time of sleep, deep sleep, and shallow sleep duration even though mean duration of the final period was increased and decreased, respectively.
Based on system’s visualizations in the Summary One-Day section, all the participants had problems regarding their sleep in the beginning of the observational period, while in the final period, fewer problems were observed (Table 5, Fig. 12). Our participants in the beginning were more than an hour in bed awake, their total shallow sleep duration was more than 5 hours, and their Deep Sleep duration was limited. Furthermore, more than 4 interruptions during nighttime sleep activity were detected. In the final period, the clinician observed through the system decrease of total time awake in bed and total time of shallow sleep, increased duration of total time of deep sleep and total time asleep. Major problems of sleep were observed initially in our third participant, while there was absence of REM sleep activity during the night and decreased duration of deep sleep, but significant improvement was observed in the final observational period Figs. 13 and 14.
Furthermore, as said through the interface, the clinician was able to define specific sleep parameters and set thresholds in the Dashboard section for each sleep pattern. For instance, sleep duration: 7 hours, number of interruptions: (more than) 2, sleep latency: (more than) 30 minutes, days of reoccurring problem: 3.Thus, regarding our first participant, through the interface, it was revealed that there was a decrease in the number of interruptions, short sleep duration, and total time awake in bed (March-April: 74 problems detected; May: 74 problems detected; and June-July: 23 problems) (Fig. 15). For our second participant, there was an obvious decrease of number of interruptions, short sleep duration, and total time awake in bed (April: 99 problems detected; July-August: 88 problems detected; and September-October: 50 problems). As for our third participant, improvement was actually observed in short sleep duration and total time awake in bed (Aug-Oct: 48 problems detected; Nov: 25 problems detected; and Dec: 7 problems). Finally, regarding the fourth participant, the clinician noticed improvement in short sleep duration and total time awake in bed (Oct: 32 problems detected; Nov: 24 problems detected; and Dec: 15 problems).
Physical activity and activities of daily living
In Table 6, the highest values of moving intensity as presented in Summary One-Day section from initial and final period are shown for each participant. Graphs for a specific period highlight problems regarding moving intensity in the initial period and significant improvement after adaptive clinical interventions (e.g., exercise programs). Most of the values in the Daily graphs of moving intensity for our first participant were very low in the beginning, while after the intervention program and installation procedure, were increased. The statistical analysis also confirmed the significant improvement over the time regarding moving intensity and ADLs for all participants.
Regarding the first participant, even though there was an improvement to all aforementioned domains, statistical significant performance was found in kitchen presence while the participant spent many hours in the kitchen for meal preparation and cleaning after the observational period (M = 2.40, SD = 1.41) than in the initial (M = 1.54, SD = 0.62); t (63) = 4.63, p = 0.000, whereas no statistical significant differences were found regarding moving intensity, washing machine, TV usage, and iron use.
As far as the second participant is concerned regarding bathroom presence, a statistical significant difference was found because the participant visited the bathroom less after the observational period (M = 14.01 min, SD = 1.51) than in the initial (M = 19.9 min, SD = 9.36); t (37) = 2.19, p = 0.03. Also, less TV use after the observational period (but not statistical significant) was detected, while moving intensity remained stable after the observational period.
There was a significant improvement in moving intensity after the observational period (M = 74.25, SD = 40.8) than in the initial (M = 68.62, SD = 39.87); t (1443) = –3.7, p = 0.000 in the third participant. An increase (but not statistical significant) in kitchen presence was noted after the observational period, and significantly less TV use after the observational period (M = 9.29, SD = 4.25) than in the initial (M = 11.05, SD = 4.09); t (51) = –2.1, p = 0.03, was recorded.
Finally, in the fourth participant, TV use was increased use after the observational period (M = 11.4, SD = 4.19) than in the initial (M = 4.13, SD = 4.42); t (31) = –6.11, p = 0.000; however, an increase in bathroom presence and taking care of her personal hygiene was observed after the observational period (M = 32.1, SD = 22.7) than in the initial (M = 13.2, SD = 9.77); t (22) = –3.91, p = 0.001. For moving intensity, no statistical significant difference was found.
Correlations between metrics
The most significant and important correlations between sleep and moving intensity or betweendifferent sleep patterns of our participants are presented in the following figures based on thevisualizations of the Comparison per Day or per Week sections. It is clear that after an intensive activity, the duration of sleep was increased in our fourth participant and when limited moving intensity was detected, the duration of deep sleep was also low (arrows) (Fig. 16). Moreover, in our second participant, we could see that the more the participant spent in bed awake until he fell asleep (total time in bed awake), the more interruptions during the night were detected (Fig. 17). Correlation between moving intensity and total time asleep provided a rich set of quality information about our third participant with cognitive impairment as well (Fig. 18), similarly to our first participant, where expected results and negative correlations observed regarding moving intensity and sleep latency (Fig. 19).
DISCUSSION
In general, the proposed system is able to provide all the necessary tools to clinicians in order to efficiently support the participants and promote their quality of life via assistive technology by focusing on practical aspects of everyday activities. Even though many studies have been conducted about how to technologically support people with dementia in order to complete successfully daily tasks, the novelty of the proposed system is that it allows different combinations and correlations of sensors to be selected based on the clinical needs of every participant. Moreover, the system identifies potentially problematic areas across several domains (sleep, physical activity, ADLs) by using individualized problem-detection parameters, and examines these patterns to identify improvement, stasis, and deterioration over time. Also, the system provides objective data for functional domains to the clinician through visualizations and monitors sleep stages during night time activity. Additionally, adapted and personalized interventions based on regular sensor-based monitoring, which combined with automatically or manually generated reminders found to improve clinical status and eliminate cognitive deficits of people with cognitive impairment as shown mainly by the clinician interface. Also, a key strength of the proposed system was that through the clinician interface trends it could be assumed how participants and caregivers integrated the technology into their everyday life and how this helped them to face and address their real needs and problems. Those interventions aimed to tackle those difficulties in cognitive function, physical status, and performance of daily activities as observed by the clinician through the system. Many studies have shown that clinical interventions in the homes of elders slow down and forestall functional decline compared to home care without such interventions (e.g., nursing, case worker visits) [57]. Similarly the clinician could also see the potential benefits of the sensors for monitoring sleep, physical activity, IADLs, and mood as this would help inform interventions and track progress. However, success would be dependent on those factors outlined above; PwD’s goals, design and functionality of the sensors, PwD’s learning capacity, levels of learning support available, and the person’s attitude toward technology. During the observational period, it became obvious that people at different stages of dementia had various needs and as a result they had different demands for clinical interventions and cognitive rehabilitation strategies. Thus, the installation procedure and personalized interventions were different for every participant.
After the installation of the system and the specific interventions, we discovered that our participants improved in cognitive function, ADLs, and many behavioral aspects. They became more aware about their personal issues and problems, and we noticed significant improvement regarding nighttime sleep. These positive results are mainly attributed to the system for the following reasons: a) early detection of problems or issues that could not be identified through clinical assessment only, b) objective and regular measurements, c) successful personalizedinterventions, and d) direct guidelines from the system and the clinician to the participant.
Firstly, in all participants we observed improvement in their neuropsychological examination, even though two of them were at a mild stage of dementia. Finally, at the end of the intervention, there was significant improvement in various tests and particular in in the MMSE, MoCA, RBMT, Hamilton Depression Rating Scale, and Rey Auditory Verbal Learning Test, which is a very promising evidence if we consider that the majority of the studies have highlighted the progressive cognitive deterioration in most of the participants with cognitive deficits [58–60]. These results indicate that cognitive deficits can be ameliorated after specific adaptive non-pharmacological interventions while improvement in memory and learning can be observed in people with cognitive impairment. These cases suggest ways in which an information and communication technology solution in correlation with psychosocial interventions can be adapted to accommodate the limitations in comprehension, memory, learning, attention, and night sleep that are common in individuals with cognitive disorders. For example, tailoring simplified checklists and messages from clinician through End-users interface, which mainly require recognition skills and can be used as reminders, were found to be very helpful for self-monitoring and/or practice exercises. Also, keeping a notebook with monitoring forms in the same place at the participant’s residence, facilitated documentation of homework exercises.
All participants described here had clinically meaningful reductions in anxiety and improvement in their cognitive performance after the installation of the proposed system and after their participation in aforementioned clinical interventions. This suggests that clinical interventions followed by the proposed system may be valuable for reducing levels of anxiety, improving cognition, ameliorating sleep problems, and promoting independence in performance of IADLs in individuals with cognitive deterioration. Relaxation exercises, CBT-AD, weekly meetings with a clinician, increasing activity, and other approaches may be especially useful. In addition, all participants reported benefiting though interventions. In general, this study found that memory training, exercise and daily activity, and psychotherapy can improve the cognitive limitations of people with cognitive deterioration and release their anxiety and behavioral disturbances. Even though this study revealed positive effects of memory training, every participant had to receive 40 minutes of cognitive rehabilitation exercises and memory training, two times per week for a total of 15 weeks (except participant 4 who received interventions for 11 weeks).
Furthermore, the interoperability of the interface helped the clinician to make many correlations between multiple variables and was found to be very useful for the clinician’s observations and decision making. Based on the system’s output, there was improvement in the frequency and duration of ADLs of all participants. Moreover, the system provided evidence regarding less TV usage in the participants’ daily lives, which is a hallmark of aging well and improvement in quality of their life. Not only could the clinician detect cognitive changes, but behavioral disturbances such as depressive and anxiety symptoms also seemed to fade away during the observational period and follow-up calls and visits by the clinician. For instance, a common problem observed in participants was the limited daily physical activity and moving intensity. However, it is widely known that participants with cognitive impairment usually demonstrate reduced levels of daily physical activity in contrast with healthy adults, because they find it difficult to come up successfully with daily activities. In a recent study [16], it was shown that changes in mobility patterns monitored by motion sensors were related to changes in cognitive ability. Similarly to our study, our participants initially had limited daily physical activity. Nevertheless, when adaptive interventions, which targeted increasing physical activity, were introduced we found significant improvement regarding participants’ daily moving intensity. They attended specific exercise programs for elderly, and during the meetings with the clinician the improvement was obvious. For instance, the participants were looking after their home (cleaning, use of washing machine, ironing), themselves (showering), and taking care of medication (right time and dose of each drug) without assistance. This change had a positive influence on the participants as was detected from the system (wearable sensor-moving intensity levels). Moreover, all participants reacted positively to engaging in these exercise activities while their physical discomfort decreased. This finding shows that daily activity is a promising nonpharmacological approach particularly relevant to the rehabilitation context [61]. Also our results indicate that the consistently positive benefits of the therapeutic use of activities are in line with other studies that have deployed similar intervention program in nursing home residents [62] and have found that activity engagement is an important source of meaning to patients, providing a positive emotional outlet and sense of involvement and belonging. Our results also show that non-pharmaceutical interventions such as reminiscence have been identified as potentially effective, especially for management of life satisfaction, mood, depression, cognitive function, behavior, memory, and well-being as has already been proved from our previous study [46] and others as well [45, 63–65]. During the final period, all participants were using electrical appliances correctly in their home and taking care of their personal hygiene more intensively and frequently than in the beginning. After clinical guidelines and advice, all participants became more careful about housekeeping. Regarding our fourth participant, who was facing problems in remembering her medication, we observed significant improvement as well both from the clinician interface and as her son noticed in the caregiver interface too. This finding shows that cognitive rehabilitation strategies, which use cues and prompts, helped the participant to improve daily functionality and engage them in daily activities (e.g., cooking). It is helpful to add visual cues, followed by tactile cueing if necessary. It proved to be very important to use very specific prompting that clearly instructs the individual to perform the specific steps of anactivity.
The system provided clinicians with very useful information regarding participants’ sleep qualities at night (total time of sleep, number of interruptions, etc.). In particular, sleep disorders and interruptions during nighttime is a common characteristic in the majority of types of dementia, especially in AD [66] and other neurodegenerative diseases [67], but to the best of our knowledge, there is not any research that has shed light into that field of home monitoring of sleep activity via visualizations of people with dementia. Similarly, our participants who met the criteria of MCI and dementia declared initially that they face problems with their sleep at night. More specifically, regarding our first participant, in the beginning of the observational period, she had very intense sleep problems, while in the final period there was a clear improvement. Based on the figures and graphs of the interface, it was determined that specific measurements of sleep had intense problems and based on these visualizations, specific interventions were introduced. The second participant also experienced major problems with his sleep based on the sleep sensor output and participant’s statements. Through the Clinician interface, the clinician identified abnormal activity of REM sleep, a common characteristic in neurodegenerative diseases [68–70] and which helped the clinicians to detect the early onset of primary supranuclear palsy (PSP). Our data also suggested that the mechanism underlying excessive daytime sleepiness existed in our PSP participant. The study of the sleep stages and specifically REM activity has been used as contributors to support the diagnosis of PSP patients and have been investigated only via neuroimaging (e.g., EEG) methods until recently. Our research shed light to a new field of studying various sleep patterns via a sensor-based home monitoring system embedded in participants’ home environment where no other confounding factors (e.g., anxiety due to unfamiliar laboratory environment) can affect the results. Moreover, regarding our third participant, she had intense problems with her sleep both in short sleep duration and total time spent awake in bed, but following the adaptive interventions, the participant demonstrated significant changes in her sleep during the final observational period. Finally, for our fourth participant, we noticed improvement both in sleep latency duration and number of interruptions during the nighttime sleep activity. This is a very promising result due to the fact that sleep behavior disorder is consider as a hallmark for disease progression and a risk factor for AD. The results of this study support the use of a remote monitoring system of elderly subjects for sleep and related problems associated with dementia, and changes in functional capacity. Consequently, the intervention results showed that the anxiety and depressive symptoms were significantly improved, while sleep disturbances and interruptions during nighttime sleep activity decreased after application of CBT-AD therapy, systematic relaxation technique, and group psychotherapy. This finding is consistent with that of the study suggesting that CBT-AD can improve anxiety in a small number group of patients with dementia [47]. It must be stressed that behavioral limitations affecting the dailyfunctionality (e.g., sleep) and cognitive function (e.g., memory) of patients with dementia as detected by the clinician interface can lead to conflicts between the patients and with others, as noticed from weekly meeting with the clinician. The results of this study suggest that the reason for the improvement of these specific behaviors was the relevant adaptive interventions, which were based on the system’s modalities seen by the clinician.
Furthermore, the proposed system provided our participants and their families with relevant information about their health and lifestyle. They in turn became more knowledgeable and aware of their health condition, and better equipped to safely assume responsibility for their own self-care. In general, it is known that caregivers struggle to help people with dementia with IADLs. With such demands, it is very important to improve the quality of life of these participants and provide remote managing of the participant from caregivers [6]. Indeed, many previous studies are aimed toward providing independence and confidence for PwD and safety for their caregivers but only with methods, which show them the exact location of the patient. In our study, regarding the user interface, one of the most important aspects were the messages (prompts, reminders, guidelines) that the participant or the caregiver was able to see through a specific mobile tablet device. The user interface proved to be a better solution for the fourth participant’s problems with medication. At the outset of the intervention, the participant rated her ability to manage her medication. By the end of the intervention she rated her ability to manage medication as good/excellent and reported that she was now extremely satisfied with her ability to have the control of it. Similarly to a recent systematic review, it has been found that digital prompts and messages are very helpful to many mental diseases [71], but it needs to be stressed that there is no other system which has provided participants or caregivers with information, simultaneously, regarding their sleep, daily activity, and physical condition through simple graphs and understandable visualizations. Finally, the caregivers were able to monitor the progress, the issues, or even the problems that the End-user & caregiver interface provided to them. Contrary to other studies that have found that IT applications are difficult to use, and are not attractive [72] or make participants feel like “machines, which are being instructed” [22], in our research we found that our participants and their caregivers found the End-user & caregiver interface very easy to use, attractive, unobtrusive, and helpful in their everyday life [73]. The acceptability of the system was extremely highlighted and the integration of it into their daily lives was very important as shown by participants’ statements:
“ I had a major problem with sleep. I noticed that if I was worried about something I would wake up a few times and could not sleep again. After all these interventions [supported by the system] I could sleep better. I could notice the improvement myself by tracking my sleep on the tablet. It feels like I am kind of “in control”. In control of my day.” [A.V., 1st participant].
“ To be honest in the beginning I couldn’t believe that my father’s condition could improve in such a short period. I was about to hire a full-time caregiver for him [ ... ] I cannot describe with words exactly the [comforting] feeling of being at work in the morning and have full access all day, from the tablet, to see how he slept, what he did, if he is ok etc. It is a feeling of safety and relief that every caregiver of an elder must have.” [P.K., 2nd participant’s caregiver]
“ My involvement in this program came the exact time when I desperately needed help, due to my major problem with my memory. I was scared that I was losing my mind. Now I am well. [During the program] I was tracking my progress by myself on the tablet. [ ... ] I used to be a physician but I couldn’t imagine that I would take part in my own health management and treatment without actual medication. This works.” [V.T., 3rd participant]
“ In the beginning I was doubtful and double-checking to make sure that the system’s indications about my mother’s behavior and symptoms are correct. I would double check for actual sleep deprivation and stress signs, then I saw from the system that indeed they were there. I realized that this was the solution I was looking for. It would help me solve my problems. Then the clinical interventions took place and she actually started remembering things like taking pills in a regular basis. I feel very safe and sound now.” [V.Z., 4th participant’s caregiver].
Limitations
A number of potential areas for improvement were identified from the evaluations of the system. For instance, retrospective data analysis for the clinician as patterns or points of change can be identified, but the participants find it difficult to remember life events that coincided with these times. To address this issue, we encouraged our participants to add notes about their day to evaluate the sensor output, although people with mild to moderate stages of dementia forgot to complete it on a daily basis. The importance of context in understanding anomalies or points of behavior change has been identified in previous research. This reporting should not be mandatory or required each day; instead it should enable key events to be noted. Finally, as new technology becomes available, the system can be updated with new sensors offering more modalities, such as skin conductance, and maintaining the same or even decreasing the level of unobtrusiveness for the end user with respect to size, comfort, and battery life, even for those participants familiar with technology. Also, for people with milder levels of impairment, areas highlighted as important involved the promotion of functional independence through continued active participation in the activities of daily life, community and social participation, aiding communication and memory, stress reduction, and encouraging physical activity. For people with moderate levels of impairment, the focus changed to more basic areas or to areas involving close personal contact with a single person. This heterogeneity could be addressed by selecting only participants at the same stage of cognitive impairment but in this study we would like to find out how PwD have different needs and if the proposed system can fulfill them.
Conclusions
People affected by dementia experience a progressive decline of their cognitive functions. As their independence diminishes over the years, the burden of their caregivers gradually increases. The lack of the public services to offer a better quality of life and the limited access to cognitive rehabilitation programs, highlight the demand for assistive technology to promote care, improvement, and facilitate accurate clinical assessment and personalized interventions. Flexible visualizations can be combined with participants’ reports and from caregivers to improve the understanding of everyday life for the PwD, gaining new insights into difficulties and demands that affect quality of life, and better assist individuals in completing daily activities while maintaining independence. Positive results in four home pilots were consistent: physical activity levels and sleep quality, as monitored through the system, lead to non-pharmaceutical interventions and improvement, leading to better cognitive status for all participants. We conclude that the proposed system with adaptive interventions is potentially useful in treating dementia participants and that this technique merits further study.
Our research describes the participation of persons with aMCI and dementia in the development of an integrated multimodal assistive technology solution to support community dwelling people with mild to moderate cognitive impairment in their daily life. Thus, we believe that strengths of our study are that different types of participants were included (people with MCI and dementia along with their caregivers in some cases) and that they participated in several phases of the development process in order to apply adaptive personalized interventions. Our study contributes to scientific research by providing insight into the development of technological solutions that are adapted to the needs and wishes of various target groups (from mild to severe stage of dementia). Focus groups, interviews, and testing with a prototype of the assistive technology to be developed proved to be useful methods to collect relevant data that could be used for further development of assistive technology. For the abovementioned reasons, our selection of two different types of cognitive impairment (MCI and AD) was on purpose since we wanted to test different aspects of the dementia spectrum and make our system a personalized solution for different types. Finally, the statistical analysis conducted showed that our clinical observations are meaningful and significant.
Further surveys concerning the adjustment of our system in participants at different stages of dementia are needed, and should be realized during future larger scale deployments as we now have a system that is stable, and quickly deployable and adaptable to different environments. Still a significant contribution is provided in this paper, as we demonstrated the system’s efficiency to detect signs of cognitive deterioration and abnormal behaviors of participants and apply specific interventions based on system’s output. Our future plan is to include more people, in collaboration with Alzheimer Hellas, with other types of dementia (i.e., Lewy body dementia, frontotemporal dementia, etc.) and cognitively impaired patients at various stages of disease (i.e., subjective cognitive impairment, severe dementia, etc.). We are currently running two installations for a period of one year in two homes of people with amnestic and multi-domain MCI diagnosis.
