Identification of ideal contexts to issue reminders for persons with dementia

Abstract

Dementia is a global health concern that primarily effects cognitive functioning, leading to forgetfulness and reducing the capacity for independent living. In this paper, we present an app designed as a reminding aid for persons with dementia as part of a 12-month randomised control trial with 125 participants who have shown a decline in cognition, evaluated by a Modified Mini- Mental State Exam (TC cohort). The app was also evaluated by healthy adults from the University of Ulster (HC cohort). In addition to reminding, the app also acts as a sensor data collection tool, which records selective data from a range of sensors around the time a reminder is delivered in an effort to gain an insight into relevant contextual information. To date, over 3000 sensor recordings from both cohorts have been collected and analysed. The recordings have been used to develop and validate a model that can identify in which contexts a reminder is typically acknowledged or missed, allowing for context-aware delivery of reminders or notifications at a time when the individual is mostly likely to receive the prompt. Using data from both cohorts weakened the accuracy of the model for the TC cohort, signifying that the TC cohort require their own non-generalised model. Future work will involve implementing the models developed into the app based on the existing TC data, so that the reminder delivery can be altered in real-time for this cohort.