Falls Prediction in Care Homes Using Mobile App Data Collection

Abstract

Falls are one of the leading causes of unintentional injury related deaths in older adults. Although, falls among elderly is a well documented phenomena; falls of care homes’ residents was under-researched, mainly due to the lack of documented data. In this study, we use data from over 1,769 care homes and 68,200 residents across the UK, which is based on carers who routinely documented the residents’ activities, using the Mobile Care Monitoring mobile app over three years. This study focuses on predicting the first fall of elderly living in care homes a week ahead. We intend to predict continuously based on a time window of the last weeks. Due to the intrinsic longitudinal nature of the data and its heterogeneity, we employ the use of Temporal Abstraction and Time Intervals Related Patterns discovery, which are used as features for classification. We had designed an experiment that reflects real-life conditions to evaluate the framework. Using four weeks of observation time window performed best.