People with severe mental illness (SMI), such as schizophrenia, bipolar disorder and major depression, also suffer from much poorer physical health than the general population. Promoting healthier lifestyles can help to prevent or combat many of the common physical health problems people with SMI commonly suffer from, such as diabetes, heart disease and lung disease. Yet, the physical health needs of people with SMI have, in the past, often been overshadowed by their mental health needs. Part of the problem is that mental health services are over stretched and under resourced, leaving little time to build a detailed picture of a person’s physical health. Even when this does occur, assessment of daily activities (including physical activity, exercise and sleep) is often reliant on self-report questionnaires and one-off assessments in clinic, and is therefore subject to bias.
Wrist-worn, electronic ‘wearables’ offer an opportunity to measure personal activity levels in a discreet, accurate and low-cost way. Wearables have been highlighted as a key growth area set to impact 80% of patient care pathways in the next 20 years. If patients with SMI could be offered an assessment period wearing a suitable wearable device, this could be an efficient way to help build a picture of an individual’s typical patterns of activity in everyday life. Reports on wearables data could be taken into account when patients and healthcare professionals discuss lifestyle changes to promote physical health.
This project aims to analyse the activity patterns of people with SMI using data collected using wearables, collected as part of the UK Biobank study. Their data will also be compared to people with no mental illness. We will use data-driven, ‘machine learning’ methods try to see if people with SMI can be assigned to different groups based on their typical activity levels. We will also evaluate how useful these groupings are at summarising and describing individuals’ activity patterns, and whether they are associated with other key health outcomes (e.g. incidence of diseases and premature death). If the groupings do prove useful, we will then explore how they could be linked with recommendations to improve physical health at a group or individual level. If successful, the hope is this approach could streamline care planning, while also improving accuracy.