Abstract
ABSTRACT
The study had two core aims: (1) to identify data‐driven sleep–wake timing and 4‐week weekend catch‐up sleep ratio (WCS ratio) patterns among college students through 28‐day objective monitoring; (2) to quantify the combined associations between multidimensional sleep profiles and anxiety symptoms. A two‐phase design was used. We recruited 182 undergraduates (mean age = 19.3, SD = 1.09) for 28‐day millimetre‐wave radar sleep tracking and psychological tests. Study 1 adopted k‐means and DTW time‐series clustering, extracting two sleep–wake timing patterns (Post‐midnight,
n
= 161; Pre‐midnight,
n
= 21) and two WCS‐ratio trajectories (decreasing,
n
= 90; increasing,
n
= 92), regarded as exploratory short‐term behavioural patterns. Study 2 performed logistic regression, sensitivity tests, and cumulative indicator analysis. Key results: (1) Poor sleep quality (OR = 1.98, 95% CI [1.47, 2.66]), shorter weekday sleep duration (OR = 0.52, 95% CI [0.29, 0.92]), pre‐midnight sleep timing (OR = 9.84, 95% CI [2.01, 48.06]), an increasing WCS‐ratio pattern (OR = 3.88, 95% CI [1.14, 13.23]), and severe negative life events (OR = 6.22, 95% CI [1.24, 31.34]) predicted higher anxiety score. Sleep quality, weekday sleep duration, and life stress remained robust in sensitivity analyses; continuous sleep and WCS indicators showed no stable predictive effects. (2) Cumulative analysis found a non‐linear threshold effect: symptoms of anxiety tend to rise when multiple sleep and stress factors co‐occur. These findings suggest that monitoring multifaceted sleep‐related behaviours, especially their long‐term dynamics, may aid in identifying college students with elevated anxiety symptoms.