Activity Fragmentation and Posture-Transition Structure in Adults Aged 70-95 Years: A Participant-Level Analysis of HAR70+

Abstract Background: Accelerometer summaries of older adults commonly emphasize accumulated activity and sedentary time, but these totals do not show how movement is distributed into bouts or how often posture and locomotor states change. We examined whether movement amount, continuity, and state organization provide distinct participant-level descriptions in HAR70+. Methods: This secondary analysis used synchronized lower-back and thigh accelerometry with validated second-by-second activity labels from 18 adults aged 70--95 years during approximately 40 minutes of semi-structured activity per participant. Labels were grouped as ambulatory activity (walking, shuffling, and stair ascent or descent), standing, sitting, or lying. We calculated activity composition, transition density, normalized state entropy, median active-bout duration, and longest inactive bout for each participant. Participant-level means and 95% bootstrap confidence intervals were estimated with 5,000 resamples. Spearman correlations examined ambulatory time in relation to transition density and state entropy. Results: The recordings comprised 12.55 hours. Mean ambulatory time was 51.05% (95% CI 47.00--54.55), while sitting and lying occupied 30.53% (27.29--34.31). Participants made 4.60 state transitions per minute (4.02--5.23), and normalized state entropy averaged 0.669 (0.650--0.688). The participant-level mean of median active-bout duration was 2.68 s, whereas the longest sitting-or-lying bout averaged 4.87 min. Ambulatory time was weakly related to transition density (Spearman ρ=-0.220, p=0.381). Its relationship with state entropy was negative but imprecisely estimated (ρ=-0.447, p=0.063). Conclusions: Movement amount, bout continuity, and state-transition organization captured different properties of the HAR70+ sequences. These protocol-bound results support explicit reporting of temporal organization in older-adult wearable measurement studies, but they do not estimate habitual behavior or establish clinical meaning.