Work

Examining machine learning in a digital physical activity intervention to predict and prevent falls in older adults::protocol of a mixed methods feasibility study

Year: 2023
Type: conference-paper
Study design: Study Protocol
Abstract: Background: Falls often lead to minor and major injuries and are a risk to morbidity and mortality. Accessibility and adherence to in-person exercise programmes among older adults to prevent falls can be low (Sherrington et al., 2020). Hence, digital physical activity interventions are being designed and tested to improve how older people exercise, to improve physical and mental health and prevent fall (Stanmore, 2021). A systematic review found that artificial intelligence (AI) techniques may improve the prediction of falls among older adults in hospital or simulated settings, but community-based datasets were lacking (O’Connor et al., 2022). These could offer more accurate and up-to-date predictions of older adults at risk of falling and sustaining injuries at home or in a care home. Aims: To utilise a novel digital physical activity application called KOKU (https://kokuhealth.com/) to measure falls risk and help prevent falls among older adults in the community. Methods: A mixed methods feasibility study will recruit older people to use the KOKU app to collect exercise and falls related data which will be analysed via machine learning techniques. These algorithms will be utilised to create a prediction model of falls risk in older adults in the community. This will inform the co-design of an AI-based digital dashboard with older people to educate them about their falls risk and provide them with evidence-based strategies via the KOKU app to prevent falls. Conclusions: Overall, it could improve the prediction of falls risk and the provision of preventative fall strategies by leveraging the KOKU app, AI analytics, and participatory design to help reduce the occurrence of falls among older adults in the community.
Language: English
Open Access status: green
OpenAlex ID: W7112235481
FWCI:
Cites: 245
Cited by:
Related to: 10