Unpacking digital phenotyping data from wearables and smartphones in individuals who have experienced trauma

Increasing natural disasters due to climate change, COVID-19, and growing awareness of institutional and historical abuse has pushed trauma into the national spotlight. More than 70 per cent of adults will experience a traumatic event in their lifetime, and while many recover, others are at-risk of developing a range of mental illnesses characterised by disturbances in emotional regulation.

Digital health tools hold tremendous potential to assess and treat a range of mental health issues, and can work alongside standard psychological treatments to boost effectiveness, or used while an individual waits for treatment. Data collected from smartphones and sensors such as GPS, physiological data, and behaviour, can build a highly personalised profile of mental illness onset and recovery, and is a new area of research known as digital phenotyping.

While there is much excitement about digital phenotyping, we need new approaches to understand how to wrangle, interpret, and analyse these data types meaningfully.

We have conducted a longitudinal study using ecological momentary assessment, smartphone data such as GPS, and a large range of physiological data collected via wearables in 96 individuals who have experienced trauma and have mental illness. There are over 5 million data points to analyse, and psychiatric variables including:

• PTSD
• pain
• alcohol use
• sleep disorders
• dysregulated mood.

This project uses statistics and machine learning to investigate some of this dataset to explore associations between the different types of data collected. For example, are daily physiological stress levels related to alcohol use? What cardiovascular indicator reliably predicts self-reported pain levels?

We will draw on MDAP’s expertise in data wrangling, machine learning, novel visualisation techniques, and data analysis. Findings from this study will be used to provide rich new phenomenological research around mental illness in individuals who have experienced trauma, understanding at a more fine-grained level how mood, behaviour, and physiological interact.

This research will inform the development of future digital mental health tools that use complex data to drive their interventions. These findings will have implications for any research using wearables and sensor technology to treat mental illness. This includes ‘just in time adaptive interventions’ (JITAIs), which are delivered via smartphone and leverage real time ecological data to change behaviour and address symptoms. JITAIs are underpinned by machine learning models that will be built from this dataset.