Automated Algorithm for Accurate Waking Sitting and Physical Activity Estimates Without Diaries Using Thigh-Worn Fibion Accelerometers in 10-to 12-Year-Old Children

Thigh-worn accelerometry provides accurate waking sitting and physical activity estimates in children, provided that sleep periods are accurately excluded—a task traditionally dependent on participant-reported diaries. Automated algorithms present a promising alternative, provided they deliver estimates comparable to diary-based methods. This study evaluated Fibion Analytics, an automated algorithm designed to identify and exclude sleep/nonwear periods in thigh-worn Fibion data. Using 7-day, 24-hr accelerometry data from 368 children (age 11.6 ± 0.79 years, 39.7% boys), the aims were to (1) optimize parameter combinations (n = 3,024) for the smallest absolute mean bias and highest equivalency in sitting and moderate to vigorous physical activity (MVPA) against diary-reported waking hours (usual practice) in a 60% training sample (n = 215) and (2) validate the parameters in a 40% test sample (n = 145). In the test sample, the algorithm using the optimized parameters showed a median sensitivity of 0.94, specificity of 0.97, and kappa of.89 in detecting the waking time. Almost perfect comparative agreement (κ >.8) was achieved for 78.3% of participants and substantial comparative agreement (κ >.6) for 95%. For sitting, the mean bias was −7.2 min/day (limits of agreement: −112.8 to 98.3 min/day) and the mean absolute error was 41.1 ± 35.3 min/day. For MVPA, the mean bias was −1.9 min/day (limits of agreement: −13.1 to 9.2 min/day) and the mean absolute error was 3.4 ± 4.9 min/day. The algorithm provided equivalent estimates for sitting within ±15 min/day equivalency bounds and for MVPA within ±5 min/day equivalency bounds. Fibion Analytics provides accurate and equivalent waking sitting and MVPA estimates compared with the usual practice in 10-to 12-year-old children, though variability in sitting warrants caution for individual-level assessments.