To determine the reliability and stability of sleep (duration and quality) over time in young children using repeated accelerometer estimates of sleep.

One hundred ninety-four children wore Actical accelerometers for 5-day periods (24-hour monitoring) at 3, 4, 5, 5.5, 6.5, and 7 years of age. Sleep variables of interest (duration, onset, offset, latency, efficiency, and wake after sleep onset) were estimated using the Sadeh algorithm within a commercial data reduction program (ActiLife). Children were divided into various groups according to sleep stability, and demographic and behavioral differences were compared across groups by ANOVA.

All measures of sleep quantity and quality required 4-7 days of accelerometry to obtain acceptable reliability estimates, except morning wake time (2-4 days), and sleep latency (11-21 days). Average year-to-year correlations were only moderate for most measures (r = 0.41-0.51), but considerably higher than those observed for sleep latency, efficiency, and wake after onset (r = 0.15-0.24). Only 29 children were classified as sleep-stable over the 4 years. These children were less likely to be from ethnic minority groups (P = .017) and had higher levels of day-time physical activity (P = .032).

Sleep patterns in children are not particularly stable, showing considerable variation both within a week and across the years. Few children exhibit stable sleep patterns over time, yet characterization of these children might provide further information regarding how sleep benefits health.