Counting What Counts: Ensuring Wearable Step-Count Validity for Effective Public Health Interventions

BackgroundDaily step counts are a widely used metric in public health and clinical practice for assessing physical activity levels, particularly in older adults and individuals with chronic conditions. However, most commercial step counters rely on forward trunk acceleration, making them prone to significant inaccuracies during vertical, non-locomotive activities such as Stepping-in-Place (SIP). ObjectiveThis study evaluated the step-count accuracy of Google Fit, a commercial accelerometer-based smartphone application, compared to Ambulosono, a wearable sensor that captures joint-specific range of motion (ROM), during music-paced SIP sessions. MethodsThirty-six participants performed multiple SIP trials using a standardized, music-based protocol. Step counts were recorded concurrently using both devices. Data were analyzed using regression modeling, k-means clustering, and Bland-Altman agreement analysis to assess accuracy, cadence responsiveness, and detection consistency. ResultsGoogle Fit consistently undercounted SIP steps by 20-60%, showing weak correlation with exercise duration (R = 0.16). Ambulosono demonstrated strong correlations with cadence (r = 0.789) and duration (R = 0.97), and uniquely captured biomechanical trade-offs such as an inverse relationship between step height and cadence. Bland-Altman analysis confirmed a systematic negative bias in Google Fit output. ConclusionThese findings reveal critical limitations in commercial step counters when applied to non-forward-motion activities and highlight the advantages of ROM-based sensing for accurate and context-aware activity tracking. Ambulosonos robust performance suggests its suitability for rehabilitation, elderly care, and home-based exercise monitoring, where step accuracy is essential for meaningful health assessment.