Promotion of Poststroke Upper Extremity Usage: A Comparison of Commercial Fitness Trackers and a Research-grade Accelerometer

Objectives

To compare commercial fitness trackers and the “gold standard” research-grade accelerometer in regards to accurately quantifying upper extremity movements. To assess the feasibility of commercial fitness trackers in both clinical settings and natural, uncontrolled environments.

Design

In this observational study, participants were recruited to engage in a 2-hour initial visit, followed by a 3-day home monitoring period, and brief follow-up visit with an exit interview. Participants were asked to wear 3 sensors (ActiGraph, Fitbit, and Apple Watch) in a sleeve on both forearms. During the initial visit, participants performed activities of daily living (eg, dressing, typing, walking) in a laboratory environment.

Setting

Both laboratory and home environments were study settings.

Participants

Thirty healthy participants and 10 stroke participants were recruited for this study. Stroke participants were required to be right-side affected, right-hand dominant, and community-dwelling in order to be included in the study. Healthy participants were required to be ≥50 years and also community-dwelling.

Interventions

The intervention in this study was use of commercial fitness trackers and a research-grade accelerometer to measure upper extremity movements. For the home monitoring period, sensors are worn in sleeves for 3 days, with the exception of sleeping and showering.

Main Outcome Measures

The main outcome measures used for Apple and Fitbit were step counts and active minutes. Step count assesses the changes in acceleration direction, or up and down movements of the UE. Active minutes is a combined measurement of the duration and intensity of an activity. The outcome measures for ActiGraph are time (duration in activity counts) and magnitude (intensity). Data from ActiGraph is preprocessed through ActiLife and further analyzed via a custom-written MATLAB program based on validated methods.

Results

Preliminary analysis showed that the sensitivity of the commercial sensors varies across activities. During the fine motor activities (eg, using iPad), although ActiGraph indicated consistent small movements, Fitbit and Apple Watch showed limited changes. In other gross activities (eg, folding towels), Fitbit and Apple Watch estimated 60%-70% of the hand-use movements captured in ActiGraph.

Conclusions

Further analysis using correlations and regressions will be conducted to demonstrate the relationship between commercial and research-grade sensors in stroke. We expect to establish algorithms to transform commercial data to directly indicate daily hand use. The results may suggest an alternative for clinical use to allow clinicians and patients to utilize low-cost, widely used commercial sensors to understand and promote hand use in daily life.

Disclosures

none.