Remote Measurement Technologies (RMTs), such as smartphones and wearable devices, could improve psychological treatment outcomes in depression through more objective, accurate, and comprehensive measures of patient behaviour. The extent to which these methods can be successfully implemented into healthcare and used in treatment depends on their feasibility, acceptability, and potential clinical utility as tools to collect longitudinal data from which biomarkers of overall depression, individual symptoms, and outcomes following treatment can be derived.
In this thesis, I aim to explore the feasibility and acceptability of using RMTs to collect behavioural and clinical data from individuals undergoing psychological therapy for depression, as well as the extent to which they can serve as biomarkers of depression and recovery.
To address these aims, I conducted three studies: a systematic review of the literature on passive data and depression (Aim 1), focus groups with patients and clinicians (Aim 2), and a mixed-methods observational cohort study (Aims 3 and 4). A grounded theory approach was employed on the qualitative data obtained from the focus groups with patients and psychotherapists (total N = 22). The observational cohort study recruited 66 people attending psychological treatment for depression at Improving Access to Psychological Treatment (IAPT) services and followed them up for up to 7 months. Weekly questionnaires, speech, and cognitive tasks were delivered via web- or app-based surveys, and passive data on behaviours such as sleep, physical activity, and heart rate (HR) were gathered from smartphone sensors and a Fitbit wearable device. A mixed-methods design that included qualitative interviews with patients was used to assess feasibility. Measures of engagement related to feasibility were measured as attrition from the study and data availability or missing data from RMT. To evaluate the digital biomarker potential, or RMT, the associations between digital features obtained and total depression, individual symptoms, and outcomes after treatment (recovery or reliable improvement) were estimated using linear mixed-effects models.
The review identified promising digital signals but critical methodological shortcomings, such as small sample sizes, brief follow-up times, and a high degree of heterogeneity that precluded a quantitative synthesis of results. The study on acceptability found the main issues to revolve around technology access and use, data management, the replacement of and/or complement to human contact, the potential for cognitive support, increased self-awareness from self-monitoring, and the perceived clinical utility of RMTs. The longitudinal study showed an overall attrition rate of 40%, with more intense treatment and higher anxiety (but not depression) affecting attrition. Data availability varied between different devices (smartphone vs. wearable) and treatment status (current treatment vs. waiting list). Due to large amounts of missing data from the smartphone app (60–80%), digital biomarker identification was only possible for Fitbit and speech data. Fitbit-derived features were found to be associated with depression severity; people with more severe depression went to sleep and woke up later, engaged in less vigorous activity, and took fewer steps overall compared to those with lower severity. Individual symptoms such as insomnia, hypersomnia, decreased appetite, guilt, anhedonia, and fatigue were also picked up by digital features. Finally, RMT-derived features were associated with short-term as well as long-term changes in depression associated with recovery after psychotherapy, particularly features derived from sleep and speech data.
The results of this thesis support the feasibility and acceptability of using digital health tools in the remote monitoring of depression. The findings contribute to a better understanding of the methodological shortcomings in the field and provide suggestions for a way forward. This research also identifies potential barriers to and facilitators of the implementation of RMT in healthcare. Feasibility results show that differing engagement patterns can arise from different devices, treatment intensity, and clinical characteristics. Pending replication from larger samples, the findings show that RMT have potential as tools for detecting digital biomarkers of depression and recovery. Future research, however, must address a variety of issues pertaining to the transparency, reproducibility, and standardisation of data collection and reporting.
|Date of Award
|1 Sept 2023
|Richard Dobson (Supervisor) & Matthew Hotopf (Supervisor)