Predicting Mood Disturbance Severity with Mobile Phone Keystroke Metadata: A BiAffect Digital Phenotyping Study

Area of Science:

• Digital phenotyping
• Computational psychiatry
• Mobile health (mHealth)

Background:

• Mood disorders, including depression and mania, are prevalent and linked to significant mortality.
• Improved diagnostic and treatment tools are crucial for mood disorder management.
• Understanding deep digital phenotypes is key to developing these tools.

Purpose of the Study:

• To investigate the link between mobile phone keyboard activity and mood disturbances in individuals with bipolar disorder.
• To assess the feasibility of using passively collected keyboard metadata to predict manic and depressive symptoms.
• To correlate digital phenotypes with clinician-administered rating scales.

Main Methods:

• An 8-week within-subject study design was employed.
• Participants used a customized mobile phone keyboard that passively collected keystroke metadata.
• Weekly assessments included the Hamilton Depression Rating Scale (HDRS) and Young Mania Rating Scale (YMRS).

Main Results:

• A mixed-effects regression model predicted HDRS scores with R²=.63 (P=.01).
• An ordinary least squares linear regression model predicted YMRS scores with R²=.34 (P=.001).
• Multiple significant keyboard metadata features were identified for predicting mood states.

Conclusions:

• Mobile phone usage patterns, specifically keyboard activity, correlate with mood states in bipolar disorder.
• These findings support the feasibility of using passively collected keyboard metadata for detecting and monitoring mood disturbances.