Smartphone-based digital phenotyping of sighted non-24-h sleep-wake disorder: A 258-d continuous circadian monitoring study.

Abstract

Diagnosing Non-24-h Sleep-Wake Disorder (N24SWD) in sighted individuals remains difficult due to the need for long-term, high-resolution monitoring of circadian rhythms. While actigraphy and dim-light melatonin onset (DLMO) are considered gold standards, their extended use is often impractical and compliance-intensive. We present a 258-d continuous assessment using smartphone-based digital phenotyping, which unobtrusively inferred sleep-wake patterns from passive smartphone interaction data using the app-count algorithm and Tudor-Locke rules. Specifically, the Rhythm app passively logged time-stamped human-smartphone interaction events-screen on/off, notifications, and the application in use-rather than accelerometry, aggregated into non-overlapping 5-min epochs scored as sleep or wake by the absence or presence of interaction. Results were validated against 109 d of actigraphy and 7 d of dynamic DLMO sampling, showing strong concordance. Because this metric reflected human-smartphone interaction rather than a second accelerometer stream, the nearly identical circadian periods from smartphone (25.16 h) and actigraphy (25.11 h) indicate that interaction-based phenotyping can independently recover the long-term free-running period. Smartphone-based tracking offers a practical, low-burden alternative for long-term circadian monitoring in N24SWD.