Objective: To assess the impact of ad-hoc thresholds used for sleep outcome quantification from wearable accelerometers in Parkinson’s disease (PD).

Background: Sleep problems affect up to 90% people with PD. Sleep is important for daily functioning, and poor sleep significantly impacts health-related quality of life. Accurate monitoring of sleep problems is crucial in PD; however, traditional methods are limited by recall bias in questionnaires and sleep diaries and the burden of polysomnography (PSG). Wearable accelerometers are increasingly used for sleep assessment. However, the widely used GGIR package and its originally suggested time window (TW; 5 min) and angle (5°) thresholds have not been validated in PD, raising uncertainty about its suitability for PD-specific analyses.

Method: Data were obtained from the IDEA-FAST feasibility study participants with PD who underwent two nights of PSG while wearing an Axivity AX6 on the wrist (n=7). Raw accelerometery data were processed with GGIR, and performance was assessed by comparing non-overlapping 30-s epochs labels and derived outcomes, including sleep onset (SON), number of awakenings (NAWK), and wake after sleep onset (WASO) with PSG. A total of 180 combinations of angle (A1–50°) and TW (T1–10 min) thresholds were evaluated.

Results: Increasing the TW decreased sensitivity and increased specificity, while increasing the angle threshold increased sensitivity and decreased specificity (sensitivity range 40–93% and specificity 15–77%). Increasing both thresholds reduced the mean absolute error (MAE) of NAWK compared to PSG (MAE= 36.8 at T1A1 vs. 6.8 at T10A50). Increasing the TW increased MAE for both SON and WASO, while increasing the angle threshold reduced MAE. SON was primarily influenced by the TW threshold (MAE= 41.8min at T1A1 vs. 55.1min at T10A50), while WASO was more strongly affected by the angle threshold (MAE= 84.2min at T1A1 vs. 13.6min at T1A10).

Conclusion: GGIR showed greater sensitivity for detecting sleep than wake, and its performance was highly dependent on threshold selection. The impact of varying thresholds across the digital sleep outcomes suggests that outcome-specific thresholds may be needed in PD. These findings highlight the need for PD-specific threshold optimisation and validation in larger cohorts, particularly for use as an objective endpoint in clinical trials.

References: van Hees VT, Sabia S, Anderson KN, Denton SJ, Oliver J, Catt M, et al. (2015) A Novel, Open Access Method to Assess Sleep Duration Using a Wrist-Worn Accelerometer. PLoS ONE 10(11): e0142533. https://doi.org/10.1371/journal.pone.0142533

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/exploring-the-impact-of-parameter-selection-on-digital-sleep-outcomes-in-parkinsons-disease-results-from-the-idea-fast-study/