Using Wearable Sensors to Identify Home and Community-Based Movement Using Continuous and Straight Line Stepping Time

Abstract

Objective measurement of community participation is essential for evaluating functional recovery and intervention outcomes in clinical populations, yet current methods rely heavily on subjective self-report measures. This study developed and validated a classification model to distinguish between home- and community-based activities using stepping and lying data from activPAL devices. Twenty-four healthy participants wore activPAL 4+ monitors continuously while completing activity diaries over 7 days. A grid search optimisation approach tested threshold combinations for two stepping parameters: straight-line stepping time (SLS) and continuous stepping duration (CSD). The optimal model achieved 93.7% accuracy across 24-h periods using an SLS threshold of 26 s. The model demonstrated high precision with a median difference of just 7 min between the predicted and reported community participation time. Individual variation in model performance highlights the need for validation in diverse clinical cohorts. This represents a methodological advance in objective physical behaviour monitoring, enabling accurate classification of home and community activity from posture data. By identifying not just how much people move but where they move, the model supports more meaningful assessment of functional mobility and community participation. This can enhance clinical decision making, rehabilitation planning, and intervention evaluation. With potential for adoption in clinical pathways and public health policy, this approach addresses a key gap in measuring real-world recovery and independence.