Challenges in using mHealth data from smartphones and wearable devices to predict depression symptom severity: retrospective analysis

Background: Major Depressive Disorder (MDD) affects millions of people world-wide, leading to lower quality of life and high medical costs. Despite the existence of psychoand pharmaco-therapy, more than 50% of people with MDD do not receive timely treatment due in part to inaccurate subjective recall and variability in the symptom course. A more objective and frequent monitoring of mental health status may not only improve on subjective recall but help guide treatment selection. Attempts have been made to explore the relationship between measures of depression and passive digital phenotypes (features) extracted from smartphones and wearables devices to remotely evaluate well-being and continuously monitor changes in symptomatology with varying degrees of success. A number of challenges exist for the analysis of this data, however. These include: (i) maintaining participant engagement over extended time periods and therefore understanding what constitutes an acceptable threshold of missing data; (ii) distinguishing between the cross-sectional and longitudinal relationships for different features to determine their utility in tracking within-individual longitudinal variation or screening participants at high risks; (iii) understanding the heterogeneity with which depression manifests itself in behavioural patterns quantified by the passive features. Objective: We aim to address these three challenges to inform future work in stratified analyses. Methods: Using smartphone and wearable data collected from 479 participants with MDD from the EU IMI RADAR-CNS programme, we extracted 21 features reflecting mobility, sleep, and phone use. We investigated the impact of the number of days of available data on feature quality using intraclass correlation coefficients and Bland-Altman analysis. We then investigated the nature of the correlation between the 8-item Patient Health Questionnaire depression scale (PHQ-8; measured every 14 days) and the features using the participant-mean correlation coefficient, repeated measures correlation coefficient, and linear mixed effects model. Furthermore, we stratified the participants based on their behavioural difference, quantified by the features, between periods of high (depression) and low (no depression) PHQ-8 scores using the Gaussian mixture model. Results: We demonstrated that 8 (range: 2-12) days were needed for reliable calculation of most features in the 14-day time window. We observed that features such as sleep onset time correlated better with PHQ-8 scores cross-sectionally than longitudinally, while features such as awake duration after sleep onset correlated well with PHQ-8 longitudinally but worse cross-sectionally. Finally, we found that participants could be separated into 3 distinct clusters according to their behavioural difference between periods of depression and no depression. Conclusions: This work contributes to our understanding of how these mobile health derived features are associated with depression symptom severity to inform future work in stratified analyses.