Inferring loneliness levels in older adults from smartphones

Abstract

The number of older adult has increased significantly in most current societies. One problem that is accentuated in the stage of old age is loneliness which is a serious health risk. Therefore, new methods for early detection of this condition that make use of new non-intrusive technologies are required. Loneliness includes four main factors (family, spousal, social and existential crisis). In this paper, four predictive models to determine the level of loneliness of each factor are proposed, focusing on the activities that can be monitored using a Smartphone. Predictive models have been evaluated on basis of their accuracy, sensitivity, specificity, predictive values, type I and type II error rates. This paper also presents the results of the experimentation of the proposed approach in practice and with real users through a mobile application called '!'Vive!' that implements the predictive models.