Objectives: This study aims to compare amongst preschoolers, night sleep duration derived using the automated van Hees algorithm in GGIR (GGIR_VH) against the reference Sadeh algorithm (Actilife_SD), and subjective caregiver-reported sleep diaries against Actilife_SD.
Methods: Participants were 142 preschoolers (52.1% males), age 5.5 years, from the Growing Up in Singapore Toward healthy Outcomes (GUSTO) birth cohort study. Weeklong actigraphy data was collected via the wGT3X-BT accelerometer (worn on the non-dominant wrist) with concurrent caregiver-reported sleep diaries. Analyses were conducted to compare the automated GGIR_VH and sleep diaries against the reference Actilife_SD. Intraclass correlations were calculated to compare the agreement levels. Bland-Altman plots were used to investigate the bias in the mean differences and limits of agreement (LoA). Repeated measures of ANOVAs were used to compare mean differences.
Results: For the intraclass correlation between automated GGIR_VH and reference Actilife_SD, there was moderate agreement for the nighttime total sleep duration (r = 0.66) and poor agreement between diary and Actilife_SD (r = 0.04). Bland-Altman plots revealed a positive bias when comparing diaries against Actilife_SD, where diaries reported longer sleep duration. In contrast, there was almost no bias and smaller LoAs for the comparison between GGIR_VH and Actilife_SD. ANOVAs showed that comparisons between diary (M = 9.36, SD = 1.16) and Actilife_SD (M = 6.93, SD = 1.12); and GGIR_VH (M = 6.76, SD = 1.30) and Actilife_SD both yielded significant differences.
Conclusions: Overall, the automated GGIR_VH algorithm showed moderate agreement compared to the reference Actilife_SD. In contrast, sleep diaries overestimated sleep duration when compared to Actilife_SD.