Supplementary material for machine learning and statistical analyses of sensor data reveal variability between repeated trials in Parkinson’s disease mobility assessments

Abstract

Mobility tasks like the Timed Up and Go test (TUG), cognitive TUG (cogTUG), and walking with turns provide insight into dynamic motor control, balance, and cognitive functions affected by Parkinson’s disease (PD). We assess the test-retest reliability of these tasks in a cohort of 262 PD and 50 controls by evaluating the performance of machine learning models based on quantitative measures derived from wearable sensors, along with statistical measures. This evaluation examines total duration, subtask duration, and other quantitative measures across both trials. We show that the diagnostic accuracy of differentiating between PD and control participants decreases by a mean of 1.1% from the first to the second trial of our mobility tasks, suggesting that mobility testing can be simplified by not repeating tasks without losing diagnostic accuracy. Although the total duration remains relatively consistent between trials (intraclass correlation coefficient (ICC) = 0.62 to 0.95), there is more variability in subtask duration and sensor-derived measures, evident in the differences in machine learning model performance and statistical metrics. Our findings also show that the variability between trials differs not only between controls and participants with PD but also among groups with varying levels of PD severity. Relying solely on total task duration and conventional statistical metrics to gauge the reliability of mobility tasks may fail to reveal nuanced variations in movement captured by other quantitative measures. Additionally, the population studied should be carefully considered, as reliability results differ among and within groups based on disease severity.