Keywords

Introduction

Globally, the population is aging. It is expected that by 2050, 1 in 6 people will be over the age of 65. It is a significant increase from the current rate of 1 in 11 [1]. In Australia, the population aged 65-and-over is 1 in 7, and it is expected to rise to 1 in 5 by 2050 [2]. This increase in the aging population, accompanied by the prevalence of age-related chronic diseases, presents significant challenges to public healthcare policymakers, societies at large, and individuals themselves [3]. An increasing shortage of care staff exacerbates it due to the aged care and nursing industry’s inability to attract the workforce and the departure of existing staff from the workplace as they get older themselves. As a result, this will not only reduce access to the already limited residential care placements, but families will subsequently be expected to meet the health, social, safety, and other daily needs of their older parents. Hence, the future of aged care will increasingly move to the community or home [4]. As advances in technology and the Internet of Things (IoT) are becoming rapid and part of everyday life, it is inevitable that smart home developments will pave the way for future technologies supporting older people in the community and enable their quality of life as they age.

To ensure the health and well-being of older people in the community or home, an accurate assessment of the functional independence measures is crucial. Activities of Daily Living (ADL) is a good predictor of a wide range of health-related behaviors [5]. Hence, it is used routinely in the clinical setting and before hospital discharge, and patients return to home. Over the past 40 years, more than 43 indexes of ADL have been published to determine the fundamental functional disability status of both patients and the general, undiagnosed population [5]. The Katz ADL scale is arguably the first such instrument used in the clinical framework [6]. The Katz ADL assessment requires evaluating activities pertaining to bathing, dressing, toileting, transferring, continence and feeding. This instrument scores each activity with 1 if an older person can achieve it independently and 0 if it is dependent on assistance. Hence, the Katz ADL index scores will range from 0 to 6, indicating an older person’s ability to function as being dependent (0) to independent (6), respectively. Barthel’s ADL score [7] is very similar to Katz’s ADL score. It is used to measure the performance of ten daily living activities, including feeding, bathing, grooming, dressing, bowel control, bladder control, toilet use (hygiene), transfers, mobility, and use of stairs. The possible scores range from 0 to 20, with lower scores indicating increased disability. There are also other ADL scales to measure more sophisticated functional independence, such as the full range of activities necessary for independent living in the community [8], for stroke patients receiving in-patient rehabilitation [9], and for patients with cognitive impairment such as Alzheimer’s disease or dementia [10].

The challenge is that ADL assessment approaches in the clinical setting are still based on subjective assessments from a clinical assessor [11]. There are several drawbacks to this approach. First, it is neither representative of the individual older people’s living environment nor reflective of their daily living routines. Second, it is a subjective measurement because it relies on a clinical staff’s observation and corresponding responses from the individual older people being assessed. Hence, this approach’s reliability is neither objective nor consistent for it to be reliable [12]. There is also evidence of varied ADL-based assessments, which are likely to be skewed by different individual responses to questions. These responses could vary from routine- or memory-based answers to socially desirable ones, which could be different for each individual depending on their culture, language, and educational backgrounds [13]. Furthermore, communication barriers from those with cognitive impairment could have significant implications on achieving reliable ADL assessment. Thirdly, current assessments are clinically resource-intensive, particularly when applied in a home setting, making them impractical for long-term care of the older people or disabled populations.

To address subjective bias and reduce human resource time on home-based functional assessments, the ‘Smart Home’ concept was proposed in the 1980s and applied to support independent-living older people’s health and aging [14]. Along with the emergence of new technology in mobile computing, smart sensors, and the Internet of Things, smart home is becoming topical and relevant in home automation and assistance for health and well-being [15,16,17,18]. A few smart home-like products are emerging in the marketplace that employs motion, and movement sensors to detect daily activity patterns and the possibility of detecting falls [19, 20]. However, the wide adoption and deployment of smart homes in the senior community are still elusive. We believe there are two main reasons which have limited the smart home initiatives. First, the cumbersome use of technology stems from the lack of adequate design choices influencing their ease of use by the older community and influencing the users’ perception that technology could compromise their privacy and security. Second, the lack of utility—in the form of a personalized and objective measure of functional independence that determines the individual’s activities in their home setting reflects health and well-being. Current functional independence measures through subjective assessment, such as Katz ADL and Barthel ADL, are population-based one-fits-all models and not personalized to individual profiles. Given this state of the art, this chapter presents a design, implementation, and evaluation of a novel daily activity assessment tool, the objective ADL (OADL) aimed at older populations and achieved through fusing data from simple, non-intrusive, always-on, wireless sensors placed in a home environment called Smarter Safer Home (SSH).

This chapter is structured as follows. Section “A Smarter Safer Homes Approach” describes our SSH platform and the OADL index computed from the SSH platform’s sensor data. Section “Clinical Trial and Discussion” introduces a small observational trial of the SSH platform and demonstrates the OADL index’s effectiveness. Section “Conclusive Remarks” concludes this chapter.

A Smarter Safer Homes Approach

The CSIRO Smarter Safer Home (SSH) platform was designed as a home-based passive activity monitoring system without the need for residents’ intervention to capture their health state and potential needs for care support and services. The platform was designed to be interoperable with commercially available sensors and devices. Furthermore, the design included privacy and security considerations, ensuring informed consent of all monitoring and data collection and processing processes.

System Requirements and High-Level Architecture

To enable an appealing and acceptable platform that seamlessly integrates with the home of an older person, the SSH platform’s design challenge was to use only non-wearable, environmentally build sensors. From the consultation with the network of older individuals and their family members, it was clear that wearables could be intrusive, burdensome, and introduce anxiety. In general, the criterion for sensor inclusion in the SSH platform was for individuals to attend to their normal lifestyle while enabling the SSH platform to monitor their behavioral patterns silently and derive their activities towards health status assessment. It was also made clear that to maximize the protection of smart home residents’ privacy, the SSH platform can not use any video/audio sensors. Instead, it should use advanced machine learning techniques to analyze raw environmental sensor data and infer, for example, individual’s indoor activities of the state.

The SSH platform was designed as the outcome of multiple iteration cycles, takes advantage of wireless sensor technology and health home monitoring to provide a smart home integration that aligns with the older individual needs, and enables an engagement of informal (e.g., family) support and formal care services. SSH employs machine learning techniques that capture an individual’s profile of Activity of Daily Living from sensor data. Based on that, personal level functional independence can be determined. The platform includes a sensor-based in-home monitoring system (data collection), a cloud computing server (data analyses), and a client module (data presentation) with a tablet app, a family/care-taker portal, and a formal caregiver service provider portal. Figure 20.1 shows the architecture of the SSH platform.

Fig. 20.1
figure 1

Smarter Safer Homes platform and its three modules for client, family, and care service providers

Full size image

The Wireless Sensor Network of SSH

The SSH platform includes wireless sensors that can monitor the physical environment and daily human activities within the home. In each home, environmental sensors are deployed with minimal footprints, i.e., they are deployed seamlessly ‘invisible’ in key areas of the older residents’ home environment. In an observational trial conducted in 2013 [21], trial participants reported that they completely forgot those sensors 1 week after installation. These sensors communicate in-home activity data with a local server through the Zigbee protocol, enabling low-power, secure, and reliable data transmission [22].

The main types of environmental sensors deployed by SSH include motion, power and circuit meter, ambient temperature and humidity, contact and sleep sensors. Motion sensors detect the presence of people in its vicinity. They are passive infrared sensors installed in every room’s corner (one motion sensor per room) to monitor the location and transition within the home. Power plug sensors and circuit meter sensors are either directly plugged into power outlets or connected in the meter box to measure home appliances, ovens, and cooktops’ electrical power consumption. Ambient temperature and humidity sensors evaluate the indoor temperature and humidity periodically (one temperature and humidity sensor per room); usually, they report readings every 5 min. The contact sensor records open/close doors (including pantry, fridge, and wardrobe) and windows. Finally, sleep sensors are installed under the bed mattress to monitor sleep quality, length , and different sleep stages of individual residents. Table 20.1 shows a full list of sensors, and Fig. 20.2 illustrates a typical SSH sensor installation in a two-bedroom unit (approximately 70 m2). Sensors used in the SSH platform are all commercially available from Aeotec® [23], Fibaro® [24], and EmFit® [25].

Table 20.1 SSH sensors details
Full size table
Fig. 20.2
figure 2

An example SSH in-home sensor installation (area of around 70 m2)

Full size image

All sensors (except the power meter and circuit meter sensors) are running on batteries and can last around 8 months in general in a domestic home environment. It makes sensor installations flexible and easy because the sensors are untethered and can be positioned less intrusively and close to the activity being assessed, independent of a power source. It also benefits easy sensor maintenance. Furthermore, sensor communication generally requires little bandwidth and is relatively insensitive to latency. Energy-efficient communication protocols and event-based communication strategies can be applied, i.e., only uploading sensor data whenever an event change has been detected. In this way, the sensor’s battery life can be greatly extended.

Activity Recognition from Ambient Sensors

Raw sensor data collected at indoor environments are first uploaded to the CSIRO secure cloud sever, to be initially processed to extract activities of daily living. Figure 20.3 illustrates an example of data samples collected at home by the SSH platform, where the ID represents a unique sensor ID (Table 20.2).

Fig. 20.3
figure 3

An example of motion sensor data and mobility scores of one clinical trial participant. (a) One day’s motion sensor data in a home (b) Daily mobility score of an indepenent living resident over 9 weeks

Full size image
Table 20.2 Examples of sensor data collected in-home
Full size table

From the raw sensor data, activities are recognized by applying machine learning and pattern recognition algorithms. Specifically, the SSH platform evaluated five types of ADL, including Mobility, Hygiene, Dressing, Postural Transferring, and Meal preparation, which are included in many ADL instruments such as Katz ADL Barthel ADL [6, 7]. Specifically, Barthel ADL assessments’ mappings compared to the types of ADLs evaluated by the SSH platform are listed in Table 20.3.

Table 20.3 Mapping Barthel ADL to SSH ADL
Full size table

It is important to note that the Barthel ADL instrument evaluates whether an older individual can eat meals independently. However, due to limitations of technology and privacy violation concerns, we only monitored meal preparation instead, assuming the meal consumption was the natural next step. Table 20.4 lists types of ADL recognized by SSH, the assessment criteria, and the connected sensors. Every day for each recognized activity, we compute its own objective ADL scores, including S_mobility, S_hygiene, S_dressing, S_transferring, and S_meal, representing the individual’s daily performance in a given area. To compute an objective ADL score, we utilize the assessment criteria for a given domain (e.g., hygiene) and compare it against normal performance, i.e., performance during the baseline period, defined as 28 days after the SSH platform installation. Specifically, any objective ADL score ranging from 0 to 100 represents the likelihood of the assessed activity to be normal, with 0 meaning absolutely abnormal, i.e., very low, and unlikely activity performance such as no mobility for a whole day; and 100 meaning normal daily activity. In the subsections below, we explain the calculation of individual objective scores and the overall objective ADL (OADL) score.

Table 20.4 ADLs monitored by the SSH platform
Full size table

Mobility, Dressing, Postural Transferring (S_mobility, S_dressing, S_transferring)

These three ADLs are mainly evaluated through motion and contact sensors. From motion sensors deployed in all rooms, the in-home motion patterns can be inferred. With details from contact sensors, SSH can recognize the indoor movement activity and compute its performance scores, i.e., S_mobility. Figure 20.3a shows an example of motion sensor data from different rooms in one participant’s single day. Figure 20.3b shows 9 weeks mobility scores within a home when the home resident was in a rehabilitation period after hip replacement of the same participant. We can see some of the days in this 9-week, the participant’s mobility performance is abnormal, i.e., performance scores are less than 50, which is much lower when compared to the performance score during the baseline period participant.

Hygiene (S_hygiene)

This activity represents how well the resident maintains hygiene status, inferred through bathroom access and water usage. Hygiene-related activities can usually be recognized through significant changes in the humidity sensor readings in the bathroom. As illustrated in Fig. 20.4a, around 19:00, SSH assessed a shower activity when applying peak detection techniques in humidity sensor time-series data. Similarly, the hygiene scores can be calculated from humidity readings and bathroom access occurrences, as illustrated in Fig. 20.4b. It is interesting to notice that the hygiene was a bit low from September to October when the weather was cold but was higher towards the end of October when the weather became warmer in the trial location.

Fig. 20.4
figure 4

An example of humidity sensor data and hygiene scores of a clinical trial participant

Full size image

Meal Preparation (S_meal)

Extracting meal activity implies the use of multiple types of sensors deployed in the kitchen area. We assume that when all kitchen sensors report data within a short period, meal preparation is happening with high probability during that period. Specifically, the SSH platform applies multi-scale pattern recognition and unsupervised learning techniques to evaluate the probability of a meal preparation activity and the intensity of kitchen appliances usages as its corresponding score. Figure 20.5a shows one day’s kitchen sensor events. SSH computed a meal preparation between 11:07 and 12:55 with a 98% confidence and a meal preparation between 20:01 and 22:10 with 30% confidence. Figure 20.5b shows daily meal preparation probability scores over 9 weeks.

Fig. 20.5
figure 5

An example of kitchen sensors data and meal preparation scores of a clinical trial participant

Full size image

Computing Objective ADL in a Smart Home (OADL)

Having the daily ADL value S_mobility, S_hygiene, S_dressing, S_transferring, and S_meal computed leveraging all the environmental sensors in a smart home, the objective ADL score, i.e., OADL, can be calculated through their aggregation. The OADL represents the total health and well-being status of home residents in a day:

$$ OADL=F\left(\mathrm{S}{\_}_{mobility}+\mathrm{S}{\_}_{\mathrm{hygiene}}+\mathrm{S}{\_}_{\mathrm{dressing}}+\mathrm{S}{\_}_{\mathrm{transferring}}+\mathrm{S}{\_}_{\mathrm{meal}}\right) $$

Like the individual ADL scores (S_i), the OADL score ranges from 0 to 100, indicating the likelihood of the OADL being normal, with 0 implying very unlikely and 100 implying a normal day. Note that S_i represents the activity i, and F is a function learned from 28 days baseline data (assuming representative data for normal days). In our study, we asked participants to do a self-evaluation of their ADL during the baseline period via Barthel ADL [15]. We then apply Gaussian Process Regression to learn F with minimum regression error.

Clinical Trial and Discussions

Since 2013, the SSH platform has been deployed and trialed in Armidale, a local town of New South Wales in Australia. The study was supported by the Australian Centre for Broadband Innovation (ACBI) and CSIRO and recruited participants from a cohort of aged care facility residents who live in independent units. The study was conducted following Health and Medical Research Human Research Ethics (HREC# 12/17). The participants in the study agreed to the installation of in-home sensors and data analytics via SSH. The participants consented to 12 months of participation in the trial.

To be eligible to participate in the pilot, participants had to be aged over 70 years and have no home care arrangements. Participants with cognitive difficulties were excluded. Of those who self-selected (N = 23), 17 signed consent forms; however, three residents withdrew before the sensors were installed. Participants’ retention in longitudinal trials with the older population was problematic for reasons including morbidity, mortality, relocation, or others. Over the course of the 12 months of the study (9/2013 to 9/2014), there were further withdrawals. Eventually, we collected five complete sets of data (sensor and interview data at 3Overse 5 eligible participants were aged between 79 and 88 years (Mean 83.6 ± 3.6). There were more female (n = 4) than male participants (n = 1). Two participants listed primary or secondary school as their highest level of education, 2 had non-university certificates or diplomas, and only 1 had a university education. Three interviews of participants’ general daily routine were conducted education level after sensor installation [14]. Interviews were recorded and transcribed. Relatives or friends were present at most interviews and contributed to the discussion.

SSH App

To access the progress and summarized information derived from the SSH platform, participants are provided with a tablet and the preinstalled ‘Smarter Safer Homes’ app. The app interface was designed with independent living older people and a professional graphic design company during an earlier study [26]. The app displays the progress status of their daily activities of living and physical and social activity, vital signs (measured through clinical devices). Residents can connect to their family or care services via video conferencing services within the app. An example of the app’s dashboard reflecting the daily status of well-being is represented by the colored rays (Fig. 20.6). A full-extension, green-ray, indicates the individual’s well-being measures are within the expected range for that individual.

Fig. 20.6
figure 6

An example screen for the SSH tablet app

Full size image

In contrast, two-thirds of an amber ray (not presented within Fig. 20.6) means a decline to the unexpected range. Furthermore, a one-third red-ray implies a very unexpected well-being measure that should trigger an intervention by the care-takers, to clarify the individual’s current well-being. There are five main measurement modules in the SSH app: Health Check, Sleep, Social, Walking, and Daily Activity. The core module—Daily Activity, reflects the individual’s OADL. Besides this module, SSH also provides other modules to perform remote physiological data measurements, sleep efficiency and sleep stage monitor, social connectivity promotion, and indoor and outdoor step counting.

Family Portal

Family members and friends of older people living alone often are anxious about their welfare. The family portal allows others to gain an insight into the lives of the older resident by communicating some of the information about their everyday lives via a website. There are four levels of access that the resident can make available to family members or nominated contacts. These levels are No access, Limited, Standard, and Full details. Figure 20.7 shows the front page of a family member with full access to the smart home data.

Fig. 20.7
figure 7

Family portal available through an internet browser

Full size image

Care Service Provider Portal

The care service provider portal provides access to the SSH platform for formal caregivers, such as aged care service providers, to monitor the participant’s profile and OADL scores. The Service Provider Portal can present an individual’s OADLs over various periods (weekly or monthly). It can also display OADLs of previous days on the dashboard (Fig. 20.8) for trend checking and comparison purposes. The portal can be accessed by multidisciplinary healthcare teams engaged in an individual’s care.

Fig. 20.8
figure 8

An example screen for the SSH service provider portal

Full size image

Trial Result

We measure the similarity between daily OADL and self-reported Barthel ADL of each participant by computing their Pearson correlation coefficients. In our pilot study, the OADL aligns well with the self-reported ADL with an average Pearson coefficient (75%), indicating this novel instrument’s great potential as an effective tool for accurate state assessment of the individual. Figure 20.9 shows an example of the calculated OALD score and self-reported Barthel ADL from one trial participant.

Fig. 20.9
figure 9

An example of OADL vs. Self-reported Barthel ADL of a clinical trial participant

Full size image

The results for five trial participants who participated in the 10 months are summarized in Table 20.5.

Table 20.5 Similarity between OADL and self-reported ADL of all five trial participants
Full size table

Discussion

ADLs and Home Environment

In the SSH platform that we present in this chapter, non-wearable, non-intrusive sensors are deployed within the home to monitor older people’s behavior patterns. The SSH design choices assume that a modern home’s basic infrastructure is available, such as running water, electricity, private toilet, internet connectivity, to facilitate the data collection and processing. The home environment examines the principal place where a person lives and contributes to its quality. As shown above, in Figs. 20.3, 20.4 and 20.5, we can demonstrate a technology-supported assessment that objectively evaluates its habitant daily behavioral patterns.

Additional quality measures of the home itself could be derived for a wider assessment scope. At first, some existing SSH sensor data such as motion, temperature, and humidity could help quantify the quality of different parts of the home environment, such as temperature and humidity of the living room vs. bedroom. Additionally, the comfort quality of the home environment for the older residents could be assessed by deploying additional environmental sensors, such as a light sensor for visual comfort, a noise sensor for acoustic comfort, or air quality sensors for optimal air quality. Furthermore, given the recent climate change, SSH datasets could be leveraged to understand the external temperature changes on the residents’ home activities. It has the potential to support the design of new homes for heat mitigation in hot areas.

Additional home environment quality measures, not considered within SSH yet, may include self-reported safety or intimacy measures, i.e., whether the residents appreciate the amount of space available for them and if they have opportunities for privacy if they do not live alone. Further home environment considerations, like the quality of the building’s construction (such as roof leaking and damp), as well as the quality of the immediate neighborhood around the home, could also be self-reported. However, they are not likely to influence the algorithmic power on the OADL computation of the SSH itself. The overall assessment of the home environment’s quality, including additional metrics mentioned above, can be another future research direction of extending SSH towards a generic smart home platform.

Human Factors Influencing the Quality of Data

In the first clinical trial that demonstrated the SSH platform’s feasibility for independent living [21], tablets were provided to older participants to observe their activity profile through the SSH app towards potentially motivating and encouraging them for better self-management. Although tablets’ interactive touch screen is a natural and friendly communication channel for most people, some participants found it somewhat difficult to use, especially when required to tap the screen with their fingers. Along with the recent development in voice and facial recognition through deep learning techniques, one possible future research direction of the SSH platform could include developing and integrating social robots as a virtual assistant, leveraging voice and gesture recognition as input.

Although the SSH platform uses only environmental sensors to enable convenience and reduce the potential intrusion for residents attending to their daily life, the major limitation is that this platform is currently reliable only for a single independent person in a single home. This is the case until a reliable environmental sensor technology distinguishes multiple residents in the home. Occasionally, this requirement could be relaxed for a dual occupancy home because there may be few joint activities between the two residents. For example, in our study, we recruited two multi-residential homes: one was a mother and an adult child, the other was a couple. Indeed, we noticed a decline in the SSH data quality of these two homes.

Nevertheless, these changes did not drastically affect the OADL score computation as the mother and adult child differed in their activities and schedules. Examples of this include cooking at different times. In the case of the couple, where one has a full-time job and is away from home during the day, the other can be assumed as living independently during the daytime.

Furthermore, to allay individual concern about privacy violation and personal information disclosure in a smart home, the SSH platform has strictly been designed not to use video/audio sensors to monitor in-home activities. Sensors are located in unobtrusive positions, such as ceilings, corners, and so on. Blinking indication lights of sensors, especially those deployed in the bedroom, are masked to avoid disruption. The SSH platform only records raw sensor data identifiable only by sensor serial numbers, and no mapping details between the sensors and the home or the resident are made available. The individual personal information is stored separately on a dedicated secure server. In our study, most participants forgot that the SSH platform was present after 1 week of the installation [21].

Smart Home and Quantified Self

The current SSH platform focuses mainly on evaluating five domains that are closely related to the ability of older people to live independently. This platform can also be expanded into a self-tracking tool that supports Quantified Self and ultimately contributes to improving the Quality of Life of individuals [27]. This could be achieved by adding additional sensors (e.g., wearable sensors [28]) to the SSH platform to extend its capabilities to measure various home and outdoor activities, including intensity of physical activity (via accelerometer and heart rate) or stress levels (via heart rate). Specifically, the SSH platform could develop a digital profile of an individuals’ daily activities. By tracking and analyzing changes and trends in the digital profiles, the SSH system could facilitate individual’s self-tracking, self-experimentation, and ultimately self-management, thereby promoting their health and well-being.

Quality of Life Technologies

The SSH platform is designed to support the independent living of older people. It satisfies the definition of a Quality of Life (QoL) technology [29]. It provides an object Activity of Daily Living index as a tool to assess the health-related QoL, i.e., ADL, of older people. Additionally, it connects older people to their caregivers through the family and service provider portals and thus enables timely interventions when necessary to maintain/prevent the decline of QoL. Furthermore, it also provides information to older people to support their goal-oriented self-management, in turn facilitating the enhancement of their QoL.

Conclusive Remarks

In this chapter, we focused on quantifying the home environment as contributing to the individual’s life quality and introduced a novel objective activity of daily life (ADL) assessment through a smart home platform. This platform enables the aging individuals to engage their family/care-takers and/or aged care providers to access information about their state in real-time and support services that maintain or enhance their health and well-being. Through rich and up-to-date information about the resident’s progress of functional measures of independence and health status via the SSH platform, more informed and timely care and support intervention can be provided following individual needs.

The novelty of the SSH platform lies in its features of providing an objective and personalized measure of ADL components and scoring through a set of miniaturized non-intrusive and non-wearable sensors in the home environment; and the ability to correlate this measure with the self-reported or care-reported status of the individual’s health and well-being. The SSH platform allows for dignified aging for all older people in the community by facilitating efficient and effective aged care delivery. Consequently, the SSH platform can deliver enhancement of the quality of life to the older people and peace of mind to their families.