Managing Perceived Loneliness and Social-Isolation Levels for Older Adults: A Survey with Focus on Wearables-Based Solutions

Abstract

As an inevitable process, the number of older adults is increasing in many countries worldwide. Two of the main problems that society is being confronted with more and more, in this respect, are the inter-related aspects of feelings of loneliness and social isolation among older adults. In particular, the ongoing COVID-19 crisis and its associated restrictions have exacerbated the loneliness and social-isolation problems. This paper is first and foremost a comprehensive survey of loneliness monitoring and management solutions, from the multidisciplinary perspective of technology, gerontology, socio-psychology, and urban built environment. In addition, our paper also investigates machine learning-based technological solutions with wearable-sensor data, suitable to measure, monitor, manage, and/or diminish the levels of loneliness and social isolation, when one also considers the constraints and characteristics coming from social science, gerontology, and architecture/urban built environments points of view. Compared to the existing state of the art, our work is unique from the cross-disciplinary point of view, because our authors’ team combines the expertise from four distinct domains, i.e., gerontology, social psychology, architecture, and wireless technology in addressing the two inter-related problems of loneliness and social isolation in older adults. This work combines a cross-disciplinary survey of the literature in the four aforementioned domains with a proposed wearable-based technological solution, introduced first as a generic framework and, then, exemplified through a simple proof of concept with dummy data. As the main findings, we provide a comprehensive view on challenges and solutions in utilizing various technologies, particularly those carried by users, also known as wearables, to measure, manage, and/or diminish the social isolation and the perceived loneliness among older adults. In addition, we also summarize the identified solutions which can be used for measuring and monitoring various loneliness- and social isolation-related metrics, and we present and validate, through a simple proof-of-concept mechanism, an approach based on machine learning for predicting and estimating loneliness levels. Open research issues in this field are also discussed.

A Machine-Learning-Based Analysis of the Relationships between Loneliness Metrics and Mobility Patterns for Elderly

“…However, from the study conducted in [5] to analyze the sensor data, it was observed that the algorithms mostly used for analysing continuous data from sensor are neural networks, support vector machines [19], random forest [20] and ensemble algorithms. Additionally, in our previous work on managing perceived loneliness in [3] we presented a hypothesis using machine-learning algorithms to identify the relationship between sensor collected data and loneliness levels. The algorithms we used in [3] were logistic regression, support vector machines, random forest.…”

“…Additionally, in our previous work on managing perceived loneliness in [3] we presented a hypothesis using machine-learning algorithms to identify the relationship between sensor collected data and loneliness levels. The algorithms we used in [3] were logistic regression, support vector machines, random forest. These algorithms were applied on the continuous floating point data and the categorical data.…”

“…Moreover, loneliness feelings can be triggered by numerous factors, such as changes in the built environment, living conditions, life changes, mobility, or loss of strength or health, etc. [2][3][4].…”

“…Our previous work in [3] introduced the idea of a machine-learning based architecture for monitoring the levels of social isolation and/or perceived loneliness in the elderly with the help of wearable sensors. Through [3], hypothesis have been studied via proof-ofconcept using the dummy data which was continuous in nature and resembling sensor data. Machine learning algorithms have been used to develop model to identify the relationship between sensor data and risk of loneliness.…”

“…Moreover, this study have also identified various indoor and outdoor mobility patterns which could be useful in identifying user's behavioural context. Various wearable sensors from the market for the monitoring of loneliness, such as Pozyx system, Oura Ring, Pico MiniFinder, Moodmetric Ring, Withings ScanWatch, Imosi Smart Bracelet P11, Fitbit Luxe and Garmin Instinct were analyzed and assessed in [3] in terms of their attributes, energy consumption, obtrusiveness, and ease of data extraction and two sensor categories were identified as offering promising features in the context of loneliness and social-isolation classification and prediction: Ultra-WideBand (UWB) band Pozyx brand for the indoor location measurements and Global positioning System (GPS)-based Pico minifinder for the outdoor location measurements. In addition and considering the demography of the participants in our study (i.e., the elderly), it was necessary to select the two sensors as they offered the most convenient means for the elderly to report the current state of their mood at every point in time through the use of the push buttons.…”