Validation of Two Self-Reported Sleep Duration Measures Against Objective Estimates of Sleep Duration in a Large UK Adolescent Cohort

Introduction Self-reported sleep duration is widely used but may not provide accurate information. Correlations between objective estimates of sleep and self-reported data only demonstrate if a relationship is present but do not determine if the two measures agree. We sought to determine levels of agreement between two self-reported average weekday sleep duration measures and one objective estimate(actigraphy) of average weekday sleep duration.

Methods Students (n=642), aged 11-13 years, were recruited from eight secondary schools in the Midlands region of the UK. The Schools Sleep Habits Survey was administered to obtain self-reported weekday sleep duration (WSD). Immediately after completion, students wore wrist actigraphy (GT3X+) for 7 consecutive days whilst also completing a 7-day diary.

Results Mean weekday sleep duration+/-SD (minutes) according to self-report, sleep diary and actigraphy were 534+/-76, 527+/-44, and 448+/-47, respectively. A positive correlation was observed between self-reported WSD and actigraphy measured average WSD r=0.19, p=0.01. Average sleep diary WSD was positively correlated with average actigraphy WSD r=0.31, p<0.001. Bland-Altman plots showed a mean of 82.9 and a region of agreement between self-reported average WSD and mean actigraphy WSD of 219.3 to -53.5 with nine data points falling outside of the upper and lower bound limits. A further Bland-Altman plot revealed a mean of 76.2 and a region of agreement between average sleep diary WSD and average actigraphy-measured WSD of 183.6 to -31.2 with >95% of the data falling within the levels of agreement.

Conclusion Self-reported average WSD, as measured through sleep diaries and survey data, were significantly correlated with objectively estimated WSD in our large adolescent cohort. Both self-reported measures of average weekday sleep duration in adolescents agreed with objectively determined average WSD data with 95% of the data falling within the levels of agreement when graphically plotted. Self-reported WSD may be confidently used as a cost-efficient method in future studies.

Support: Action Medical Research provided study funding.