Validity of body-worn sensor acceleration metrics to index upper extremity function in hemiparetic stroke

Abstract

Background and purpose: Accelerometers have been used to capture real-world use of the paretic upper extremity in people with stroke. It may be possible to characterize different aspects of the recorded acceleration to gain insight about movement capabilities during task-specific behavior. These measures may be of value for guiding rehabilitation. We undertook a study to identify the acceleration characteristics that have a stable association with upper extremity function and sensitivity to within-participant fluctuations in function over multiple sessions of task-specific training.

Methods: Twenty-seven adults 6 months or more poststroke with upper extremity paresis participated. Signals from wrist-worn accelerometers were sampled at 30 Hz during 7 sessions of task-specific training. Paretic upper extremity function was evaluated with the Action Research Arm Test. We used Spearman correlations to examine within-session associations between acceleration metrics and Action Research Arm Test performance. A mixed model was used to determine which metrics were sensitive to within-participant fluctuations in upper extremity function across the 7 training sessions.

Results: Upper extremity function correlated with bilateral acceleration variability and use ratio during 5 and 6 sessions, respectively. Time accelerating between 76% and 100% of peak acceleration correlated with function in 6 sessions. Variability of the paretic upper extremity acceleration and the ratio of acceleration variability between upper extremities were associated with function during all 7 sessions. Variability in both the acceleration of the paretic upper extremity, and acceleration of the paretic and nonparetic extremities combined were sensitive to within-participant fluctuations in function across training sessions.

Discussion and conclusions: Multiple features of the acceleration profile track with upper extremity function within and across sessions of task-specific training. It may be possible to monitor these features with accelerometers to index upper extremity function outside of clinical settings.Video Abstract available for more insights from the authors (see Supplemental Digital Content 1, http://links.lww.com/JNPT/A91).

Conflict of interest statement

Conflict of Interest: The authors declare no conflicts of interest.

Figures

Figure 1

Wrist-worn accelerometer devices (ActiGraph LLC, 49 East Chase Street, Pensacola, FL 32502; http://www.actigraphcorp.com/).

Figure 2

Accelerations in each cardinal plane for the A) paretic and B) non-paretic upper extremity during 60 seconds of task-specific training. Accelerations in each cardinal plane were combined into a composite acceleration for the C) paretic and D) non-paretic upper extremity.

Figure 3

Correspondence between actual values for acceleration metrics and ARAT score across the seven sessions of task-specific training in two example participants: A) variability of the paretic-upper extremity acceleration and B) variability of the bilateral acceleration.