Pilot Site Deployment of an IoT Solution for Older Adults’ Early Behavior Change Detection

The world demography is continuously changing. During the last decade, we noticed a regular variation in the world demography leading to a nearly balanced society share between the young and aging population. This increasing older adult population is facing many problems. In fact, the transition to the aging period is associated with physical, psychological, cognitive, and societal changes. Negative behavior changes are considered as indicators of older adults’ frailty. This is why it is important to detect such behavior changes early in order to prevent isolation, sedentary lifestyle, and even diseases, and therefore delay the frailty period. This paper exhibits a proof-of-concept pilot site deployment of an Internet of Thing (IoT) solution for the continuous monitoring and detection of older adults’ behavior changes. The objective is to help geriatricians detect sedentary lifestyle and health-related problems at an early stage.


Introduction
Aging is often related to significant changes in physical activities, mobility, nutrition, social life, and cognitive status. These changes considerably affect older adults' quality of life. According to the World Health Organization (WHO) [1], the biggest health risk for seniors is the adoption of a sedentary lifestyle that causes isolation, depression and many other diseases such as cardiovascular disease, obesity, high blood pressure, etc.
In this paper, we present our pilot site deployment of an Internet of Things (IoT) solution for the continuous monitoring and detection of older adults' behavior changes. The objective is to help geriatricians detect sedentary lifestyle and health related problems at early stage, without the need to perform classical psycho-geriatric tests that have many limitations like assessment inaccuracies and the difficulty for older adults to recall past events. The work was performed as part of the European project City4Age based on six pilot sites: Athens, Birmingham, Lecce, Madrid, Montpellier, and Singapore -each of them focusing on a specific topic : Athens (Social interactions through community centers), Birmingham (Public E-Services and digital technologies), Lecce (Daily activities in public social areas), Madrid (Mobility in the city and public transport), Montpellier (Indoor/outdoor assistive services for aging well), and Singapore (social activity and engaged community within the HDB). In this paper, we will focus on and detail the deployment performed in the pilot site of Montpellier.

Montpellier Pilot Setup
Montpellier pilot site goal is to quickly and unobtrusively detect possible aging people's behavior changes. Detected behavior changes are analyzed afterwards and confirmed by collaborating geriatricians to provide adequate intervention [34].
The pilot site is coordinated by the French National Center for Scientific Research (CRNS) and is collaborating with local authorities such as Montpellier Metropolis, the ETAPE association (Health Autonomy Pole), and healthcare professionals from Beausoleil Clinic and Saint Vincent De Paul nursing home.
The proposed solution deployed in the pilot site of Montpellier consists of a set of sensors deployed in the participants' homes and in the city as shown in Figure 1. For indoor monitoring, the pilot uses motion sensors, contact sensors, and bed sensors. Outdoor monitoring is based on Smartphones carried by participants and beacons deployed in places of interest of participants in the city. These sensors allow for collect raw data and objective information in real time, which are processed to detect behavior changes.

Recruitment and Engagement
Each of the six pilot sites followed a specific ethical approval process, related to specific countries' regulation, in order to recruit participants. For the Montpellier pilot site, the ethical process consists of submitting an application to the Institutional Review Board (Comités de Protection des Personnes: CPP).
After obtaining the ethical approval, we started the recruitment process. We have approached around 40 potential participants with the help of ETAPE association and Saint Vincent De Paul nursing home. In fact, we have presented the solution in several local events to promote the project and identify interested people to be included in the study. As shown in Figure 2, briefing and presentation sessions were organized in the Montpellier pilot site to better promote experimentation objectives, quickly launch the recruitment process, and keep recruited participants involved in the experimentation. These events allowed for having in-depth discussions with interested aging people who accepted to visit a demonstration house, see a live demonstration of the system, and get their feedback. The demonstration house highlights the unobtrusiveness of the technological solution proposed for the potential participants. Interested people also observe sensor events in real-time and examples of real data over weeks and months indicating significant changes in health status. Potential participants asked for information about employed technological solutions, real benefits of adopting them at home, and possible risks. Nineteen participants accepted and have been equipped with the City4Age solution. An initial interview with included participants allowed for collecting some indications on their social and health profiles. Participants have diverse medical and social profiles such as educational level, dependence level, habits, and health status. Table 1 summarizes the social profiles of the different participants. All recruited participants were over 60 years old and living alone (11% Divorced, 5.5% Single, and 83.5% widow). In addition, 72% of them were Female (28% were Male) and 89% receive medical or social care at home. Table 2 presents regular habits and health info of some participants. Recruited participants have diverse daily and weekly habits, and medical profiles. We were interested in some specific habits such as wake-up time in the morning, toilet entries, eating time, watching television time (preferred activity for many elderly people), family and caregivers visits, etc. First, interviews with elderly people, their family members and formal caregivers investigate health status at the starting time of the monitoring period. Initial medical profiles and health status diagnosis validate participant recruitment, notify about diverse physical and cognitive problems, and provide primary explications for future detected health changes.

Technologies, Data Collection, and Data Analysis
The system proposed in the Montpellier pilot site uses indoor and outdoor technologies to monitor daily living activities of participants. For indoor monitoring, the pilot is proposing a set of sensors (motion sensors, contact sensors, and bed sensors). Outdoor monitoring is based on beacons deployed and tagging specific places in the cities considered as participants' places of interest (e.g., bus and metro stations, cinemas, restaurants, etc.). The communication between beacons and the pilot site local server is guaranteed through a dedicated mobile application installed on the the participants' smartphones. The characteristics and usage of the sensors deployed in the pilot site is detailed in Table 3. These sensors operate discreetly and allow for collect raw data and objective information in real time indoors and outdoors in the city. The objective is to be able to accurately determine the "habitual behaviors" of people by collecting data over time. Figure 3 showcases some deployed sensors in the Montpellier pilot site. For ethical reasons, the collected data will be available on request.  We have identified relevant LEAs Based on internationally-validated geriatric references (e.g., SEGA [4], MMSE [5], GDS [6], IADL [7], AGGIR [8], and others) and distinguished four main categories that can be monitored by our system: activities of daily living, mobility, social life, and nutrition as detailed in Table 4. LEAs were recognized using the semantic modeling and reasoning system we have proposed in [35].
LEAs were quantified based on time, place, number, and duration metrics in order to produce measures. These measures were analyzed by our change detection algorithms to detect possible behavior changes (e.g., sleep interruption number increases during anxiety periods, go shopping frequency decreases because of mobility impairments, toilet entries increase due to urinary infection, and preparing meal duration increases because of cognitive problems). Table 5 introduces the used metrics with some examples while the list of measures collected in Montpellier pilot site is presented in Table 6.
The deployment of our proposed solution in France allowed for collecting around two years of real data. We have collected 310,590 of Low Elementary Actions (LEA) and 49,659 of Measures for 19 participants.

Behavior Change Detection
The goal of the system proposed in Montpellier pilot site is to detect possible behavior changes that will be analyzed and confirmed by collaborating geriatricians to provide adequate intervention. A behavior change tracker service "ChangeTracker" [36] was developed allowing for detecting changes in participants' behavior using statistical algorithms. Collected data are analyzed by our "ChangeTracker" service and presented to collaborating geriatricians from Beausoleil clinic. Possible behavior changes are detected 2 to 10 days after change occurrence. Figures 5 and 6 showcase some behavior changes detected by the "ChangeTracker" service for some participants in the pilot site. In

Intervention Process
To perform intervention, the pilot site is providing a framework with visualizations about participants activities and statistics about their daily routines and habits. The "ChangeTracker" service is integrated in this framework allowing to automatically detect possible changes that can be confirmed by the caregivers and the geriatrician. The caregivers and geriatricians can navigate the data provided by these visualizations and decide on the type and form of intervention when needed. Below are some examples of interventions decided by the medical and caregiver staff: • After detecting a decrease in outdoor and indoor activities for participant 96, nursing home stuff decided to initiate home assistance • Detecting a decrease in outdoor activities for participant 92 allowed the geriatrician to decide on the hospitalization of this participant.

•
Detecting decrease in toilet visits for participant 94 and an increase in activity level for participant 95 allowed the geriatrician to change the medical treatment for these two participants

Validation
In this paper, we will focus on the technical validation of the pilot site setup and the performance evaluation of the developed "ChangeTracker" service based on medical observations and collected health records.

Technology Validation
We have firstly validated the selected sensors and devices that have been used in the deployment. In the Montpellier pilot site, we have deployed a demonstration house as shown in Figure 7 and deployed outdoor sensors in the metropolis of Montpellier. All employed technologies, such as movement sensors, contact sensors, gateways, receivers, beacons, and Internet access points have been tested in this experimental house before being deployed. These experimentation allowed for identifying and fixing some technical problems in the technologies used.
The validation of collected data in Montpellier pilot site started with the data collected from the demonstration house. This phase allowed for validating the daily activity recognition process based on collected row data and real observations. Later, we have integrated the "ChangeTracker" service to identify possible participants' behavior changes. Different statistical, probabilistic, and machine-learning algorithms have been tested [37]. These change detection techniques do not provide the same performance in terms of change detection precision. In fact, change detection techniques do not detect the same change points for tested inferred data. Therefore, applying these techniques on real monitoring data for the same monitored participants enables effective performance evaluation. Correlating change points detected by our "ChangerTracker" service with medical observations and health records is essential to evaluate the used change detection techniques' medical relevance. For this reason, we have calculated the change detection precision of the used change detection techniques. Change detection precision refers to the percentage of detected changes with retrieved medical explication (i.e., changes related to health status based on medical correlations) from the global detected changes. Unexplained changes could be related to different reasons such as false alarms or undiagnosed health problems (e.g., participant wakes-up earlier, but does not feel this change and does not find explication). Figure 8 shows measured precision for investigated change detection techniques. Bootstrap technique provides best precision of 83.87%; i.e., bootstrap technique detects significant changes that are highly related to physical and health problems.

Detection Process Validation
The "ChangeTracker" service has been validated with the help of our collaborating geriatrician. Detected changes have been correlated with medical observation and health records to validate the used algorithms and their performances as showcased in Figure 9. The "ChangeTracker" service sends change notifications from 2 to 10 days after change occurrence. These notifications inform about change date, monitored participant ID, and means before and after detected changes.
Change notifications provide an opportunity to confirm that detected changes are really permanent and investigate possible correlations with geriatric observations. Regular review meetings with older adults, family members, and family doctors allow for accurately investigating possible causes of detected changes. The pilot site local geriatrician evaluates the medical relevance of investigated change explications, by reviewing all past detected changes and correlating them with medical records (e.g., geriatric scales, cognitive diagnosis, and prescribed treatments). Review meetings investigate multi-dimensional correlations of detected changes, such as identifying parallel decreases in activity level and time out home related to mobility impairments, and consecutive increases in sleep interruptions and toilet entries after treatment change. Figure 10 shows some results of changes detected by "ChangeTracker" service and their correlation with medical observations and participant feedback. Evaluating medical relevance of detected changes requires continuous collection of personal and medical information. Personal interviews with monitored older adults enable investigation of possible causes explaining observed changes; e.g., detected changes in activity patterns correlate with personal feelings of slowness in physical movements and trembling in fingers related to Parkinson's, and detected changes in preparing meal duration correlate with forgetting recent events related to cognitive impairments. Regular review meetings allow for accurately investigating possible causes and investigating multi-dimensional correlations. Validating data collection with older adults, family members, and family doctors provides better correlation of detected changes in monitoring data with real medical health status. Professional caregivers collect electronic health records while visiting monitored participants a couple of times per day for medicine taking, toilet entry assistance, room cleaning, and nutritional services. They report all interventions, formal and informal observations, special health events, and social habits.

Results and Performance
In total, our "ChangeTracker" service detected 340 changes for all participants with an average of 0.97 change per month. Participants show diverse changes in monitoring periods that have been correlated with diverse reasons, such as physical problems, health improvements, nutritional problems, personal changes, and social problems as detailed in Figure 11.
Health improvements justify 36.76% of detected changes; e.g., participant 171 has fewer sleep interruptions correlated with improvement in sleep habits and toilet entries. Physical health problems present main detected change cause for individual house participants with a percentage of 45.59%; e.g., participant 95 sleeps longer with fewer interruptions correlated with Parkinson freezing periods. Nutritional, personal, and social changes influence monitored health status with respective rates of 8.82%, 7.35%, and 1.47%; e.g., participant 91 receives less visits at home due to social isolation.

Health Change Detection Ontology
Based on the realized work, and the collected medical interpretations of detected changes in all monitored health change indicators (e.g., activity level, sleep interruptions, toilet entries, received visits and time out of home), we have defined a global health change detection ontology, presented in Figure 12, that can be used to identify possible cause categories and sub-categories for detected health changes. Medical interpretations have been performed using records, personal interviews, and regular review meetings with elderly people and family members. The main health change causes that have been identified are physical problems, nutritional problems, emotional problems, context changes, social problems, health improvements, cognitive problems, and death. This ontology can be later extended and populated with more observations and interpretations.

Stakeholders' Feedback
Participants show considerable interest in early health change detection and more acceptance for unobtrusive monitoring. A positive feedback was collected from the participants and the different stakeholders. Below is some feedback collected from participants, geriatricians, and caregivers: We are working with patients with Parkinson's disease. In this special disease, there are many problems concerning sleep and voiding function, and we have a solution to propose to them. However, in short consultations, we don't have time to speak about all things and we know very few things about patients' activities of daily living. We think that an unobtrusive technological solution will be interesting to help us to improve our assessment. • Caregiver: My mother is participating in the project. The system doesn't affect privacy. This is very important, and our feedback is positive. We could detect changes that correlate with my mother's health status. This was beneficial for our discussions.

Conclusions
The Montpellier pilot site was a proof-of-concept of the solution proposed by the City4Age project. The pilot site realized the necessary tasks from ethical approval, building local partnerships, and deploying the proposed solution. It allowed to put in place a mechanism for the early identification of behavior changes and the provision of adequate intervention. In total, 340 changes have been detected for all participants. These changes have been validated and classified with the help of a local geriatrician and by correlation with Medical Observations. In the Montpellier pilot site, the economic value of the solution proposed by City4Age has been identified on different aspects. First, this solution could be of great value for the geriatricians, and this was confirmed by the testimony from our local geriatrician. In fact, the technological observations provided by City4Age solution enrich their medical observation for better assessments of frailty and MCI. The solution is also valuable for nursing homes. In fact, nursing homes need to remotely receive reliable information on health status, in order to offer personalized healthcare services for older adults. It requires maintaining independent living at home and identifying older adults at risk who really require entry to nursing homes. Finally, the solution can be proposed to interested aging people through public non-profit organizations like ETAPE in Montpellier. This organization provides free services for older adults to help them adapt their living environment to their needs.