Abstract
Light at night (LAN) disrupts human circadian rhythms and impairs sleep health. However, inconsistent findings across observational studies using indoor residence-based measurements (IRBM) and real-time mobility-based measurements (RMBM) raise methodological concerns. This study investigated IRBM and RMBM of LAN exposure and their associations with sleep health, examining sex as a potential moderator. Data from 484 individuals (aged 18-65) over 1,748 nights were collected in Hong Kong, incorporating objective (actigraphy) and subjective (sleep diary) measures. Concurrent LAN exposure was quantified via LUX29TK portable light meter (IRBM) and GENEActiv wrist-worn actigraph-embedded luxmeter (RMBM). Generalized linear mixed models adjusted for sociodemographic, temporal, geographic, physical activity, health status, and lifestyle covariates. Initial analyses revealed differences between IRBM and RMBM with inconsistent effects. For LAN across astronomical night (sunset to sunrise), covariate adjustment attenuated contextual discrepancies in IRBM models. Both methods demonstrated negative associations between LAN and objective sleep duration and efficiency, with no effects on subjective sleep or sex moderation. For LAN across biological night (sleep onset to wake), RMBM exhibited instrumental error, yielding divergent associations. These findings reveal how contextual and instrumental errors between IRBM and RMBM across analysis windows explain inconsistencies in prior LAN-sleep research, informing reliable exposure quantification in sleep studies.