Validity of the ActiGraph activity monitor for individuals who walk slowly post-stroke

Background: Slow and asymmetric gait post-stroke may reduce the accuracy of accelerometers (e.g. ActiGraph [AG]) to measure activity. Objectives: To (1) determine the validity of AG step counts post-stroke; (2) develop guidelines for low frequency extension filter (LFE) use; and (3) determine the feasibility of daily accelerometer wear. Methods: Adults with (n = 33) and without stroke (n = 20) wore three devices for approximately 7 h on a single day: ankle AG, waist AG, and a reference accelerometer at the ankle (REFA). AG step counts processed with and without the LFE were compared to REFA with paired difference tests. Agreement was measured with intraclass correlation coefficients (ICC3,1). Relationships between error (AG - REFA) and motor impairment and gait performance were plotted to determine a threshold for LFE application. A feasibility questionnaire was distributed to participants to investigate the applicability of the AG in clinical populations. Results: Step counts from ankle AG in the stroke group (p = 0.53) and waist AG in the healthy group (p = 0.10) were similar to REFA. Waist AG under-counted, and ankle and waist AG with LFE over-counted steps in the stroke group (all p < 0.0001). ICC3,1 ranged from 0.70 to 0.82 (stroke) and 0.79-0.92 (healthy). Ankle AG error and stance time symmetry (stroke) were correlated (r = 0.41, p = 0.02); however, no threshold for LFE application was revealed. Ankle AG was rated very comfortable by 26/33 participants with stroke and 12/20 healthy participants. Conclusions: The AG worn at the unaffected ankle without LFE produced the most accurate step count in people with stroke. We were unable to establish guidelines for LFE use.