Identifying multimodal digital features of insomnia using an app in Hong Kong: An ecological momentary assessment study

Background

Digital phenotyping of insomnia remains underexplored, particularly in the context of depression, despite the high comorbidity between these two conditions. This study aims to investigate the associations between insomnia and multimodal features using active data collection during awake states, including facial expressions, acoustic characteristics, and language use.

Methods

A sample of 92 participants was recruited, 39 % of them presented with clinical insomnia as measured by Insomnia Severity Index. Multimodal features were extracted from video-taped mood diaries recorded for one week. We analyzed the associations between multimodal features and insomnia using generalized logistic regression models, while controlling for demographic data, psychiatric diagnosis, scores of depression and anxiety.

Results

Insomnia was associated with facial, acoustic, and linguistic features, included less lip corner pulling, more upper lip raising and lip corner depressing, slower articulation rate, increased non-fluencies and the use of fillers, fewer family- and health-related words.

Conclusions

The current findings enhance our understanding of the multimodal digital phenotypic characteristics of insomnia, facilitating a more objective assessment in future research.