Reliability of IMU-based balance assessment in clinical stroke rehabilitation

Balance is often affected after stroke, severely impacting activities of daily life. Conventional testing methods to assess balance provide limited information, as they are subjected to floor and ceiling effects. Instrumented tests, for instance using inertial measurement units, offer a feasible and promising alternative.

We examined whether postural sway can reliably be measured in sitting and standing balance in people after stroke in clinical rehabilitation using a single inertial measurement unit. Additionally, we assessed to what extent averaging two measurements would improve test-retest reliability compared to a single measurement, and if sway features can potentially be used to monitor progression.

Forty participants performed two assessments with a test-retest interval of 24 h. Each assessment consisted of one sitting and four standing balance conditions (eyes open, feet together, eyes closed and foam). The standing balance conditions were performed twice during both assessments. In total, 35 sway features were calculated for each condition. For the standing balance conditions, these were calculated for both single test-retest measurement and the average of the two test and retest measurements. We determined the reliability using the intraclass correlation coefficient for both single and averaged measurements. Additionally, the minimal detectable change and the relative minimal detectable change were computed.

The single and averaged measurements resulted in 22 sitting, 30 & 32 eyes open, 27 & 22 feet together, 28 & 33 eyes closed and 23 & 13 foam sway features with good-excellent reliability. Overall, the difference between intraclass correlation coefficient values of the single and averaged measurements was small and inconsistent. The relative minimal detectable change ranged between 0.5 and 1.5 standard deviation.

Sitting and standing balance can reliably be assessed in people after stroke in clinical rehabilitation with a single measurement using one inertial measurement unit.