Principal component analysis of pre-sleep artificial light at night exposure patterns and the associations with sleep disturbances in Hong Kong shift-working nurses and daytime office workers.

Abstract

This study investigated actigraphy-measured pre-sleep artificial light at night (ALAN) exposure patterns on non-workdays and their associations with sleep parameters among Hong Kong shift-working nurses and daytime office workers. Using a 7-day actigraphy-measured longitudinal study design, we recruited 136 shift nurses and 66 matched daytime office workers (July 2022-February 2023). ALAN exposure metrics and sleep parameters were derived from actigraphy data, supplemented by sociodemographic surveys and sleep-activity diaries. Principal component analysis (PCA) was performed on log-transformed light values to identify ALAN exposure patterns; mixed-effect models analysed associations with total sleep time (TST), sleep onset latency, wake after sleep onset (WASO) and sleep timing. Shift nurses were younger (34.8 vs 36.0 years), showing no significant difference in TST (6.6 vs 6.8 hours, p=0.447) compared with daytime office workers. PCA identified Light PC1 (overall intensity) and Light PC2 (temporal pattern) patterns. In the adjusted models, wrist-worn Light PC1 showed strong associations with shorter TST (β=-0.59, SE=0.08, p<0.001), reduced WASO (β=-0.20, SE=0.04, p<0.001), later sleep onset (β=0.20, SE=0.07, p=0.003) and earlier wake-up time (β=-0.65, SE=0.09, p<0.001). Bedroom Light PC1 showed similar but weaker associations (TST: β=-0.21, SE=0.09, p=0.028). Light PC2 was not significantly associated with any sleep outcomes. Sensitivity analyses confirmed PCA's model fit, and subgroup analyses revealed no significant interaction effects (all p>0.05). Our findings suggest sleep interventions should prioritise reducing personal light exposure before bed. Future studies could clarify the roles of specific light properties.