Estimation bias and agreement limits between two common self-report methods of habitual sleep duration in epidemiological surveys

Abstract Accurate measurement of habitual sleep duration (HSD) is crucial for understanding the relationship between sleep and health. This study aimed to assess the bias and agreement limits between two commonly used short HSD self-report methods, considering sleep quality (SQ) and social jetlag (SJL) as potential predictors of bias. Data from 10,268 participants in the International COVID Sleep Study-II (ICOSS-II) were used. Method-Self and Method-MCTQ were compared. Method-Self involved a single question about average nightly sleep duration (HSDself), while Method-MCTQ estimated HSD from reported sleep times on workdays (HSDMCTQwork) and free days (HSDMCTQfree). Sleep quality was evaluated using a Likert scale and the Insomnia Severity Index (ISI) to explore its influence on estimation bias. HSDself was on average 42.41 ± 67.42 min lower than HSDMCTQweek, with an agreement range within ± 133 min. The bias and agreement range between methods increased with poorer SQ. HSDMCTQwork showed less bias and better agreement with HSDself compared to HSDMCTQfree. Sleep duration irregularity was − 43.35 ± 78.26 min on average. Subjective sleep quality predicted a significant proportion of variance in HSDself and estimation bias. The two methods showed very poor agreement and a significant systematic bias, both worsening with poorer SQ. Method-MCTQ considered sleep intervals without adjusting for SQ issues such as wakefulness after sleep onset but accounted for sleep irregularity and sleeping in on free days, while Method-Self reflected respondents’ interpretation of their sleep, focusing on their sleep on workdays. Including an SQ-related question in surveys may help bidirectionally adjust the possible bias and enhance the accuracy of sleep-health studies.