Abstract
<title>Abstract</title> <p> <bold>Objective</bold> To examine whether reported loneliness differs by survey response time and whether this association varies across demographic, socioeconomic, health-related and social-network subgroups. <bold>Design</bold> Cross-sectional observational study using time-stamped survey responses and harmonic logistic regression modelling. Exploratory subgroup analyses based on socioeconomic, household, health-related and social-network characteristics were introduced. <bold>Setting</bold> United Kingdom; time-stamped survey data from the Measuring Loneliness in the UK (INTERACT) study. <bold>Participants</bold> The baseline analytic sample included 125,963 respondents with complete loneliness and survey timing data. Adjusted analyses including demographic, socioeconomic, health-related, and social-network covariates were conducted in 19,826 respondents with complete covariate information. <bold>Main outcome measures</bold> Loneliness was measured using the ONS Direct Measure of Loneliness (DMOL) and the UCLA-3 loneliness scale. Survey response time was modelled using 24-hour harmonic sine and cosine terms, with hour-level, minute-level and subgroup-specific analyses. <bold>Results</bold> Reported loneliness varied by survey response time across both loneliness measures. Hour- and minute-level analyses showed higher proportions among responses submitted from the late afternoon onwards. In harmonic regression models, 24-hour time terms were jointly significant in baseline and adjusted analyses. Adjusted predicted probabilities were lowest at approximately 07:00 and highest at 19:00 for DMOL, and lowest at 05:00 and highest at 17:00 for UCLA-3. Additional subgroup analyses conducted to explore whether the temporal pattern of reported loneliness differed across respondent characteristics were interpreted in light of their exploratory nature. <bold>Conclusions</bold> Reported loneliness varied by survey response time, with lower levels among early-morning respondents and higher levels among late-afternoon and evening respondents. This association persisted after adjustment for demographic, socioeconomic, health-related and social-network characteristics. Survey response time may be an important consideration when analysing time-stamped loneliness data and warrants further investigation in longitudinal studies. </p>