1015 Evaluation of Actigraphy Sensors: Detecting Daytime Sleep After Stroke in an Inpatient Rehabilitation Hospital

Abstract Introduction Excessive Daytime Sleepiness (EDS) is common poststroke, significantly impacting rehabilitation participation and quality of life. Early detection of EDS through prolonged daily sleep and wake monitoring is crucial for prompt intervention and improved care. Actigraphy is a promising sleep monitoring technique that circumvents the high cost and low portability of extended polysomnography and the limited reliability of self-reporting. While actigraphy is often used in healthy adults, its application to daytime sleep measurement in the poststroke population remains unexplored. This study assesses the efficacy of actigraphy and its commonly used scoring algorithms to detect daytime sleep in the poststroke population. Methods ActiGraph wGT3X-BT and ActiWatch Spectrum were placed on the less affected wrist of participants. Trained observers monitored daytime sleep occurrences by checking on participants every 10 minutes during non-therapy periods, recording behaviors as active, sedentary, or asleep. The average observation period for each participant was approximately eight hours. Actigraphy data were cross-referenced with on-site, time-specific observations. Both the ActiWatch (Autoscore AMRI) and ActiGraph (Cole Kripke, Sadeh) use non-data-driven algorithms to estimate wake (0) or sleep (1) given a user-determined parameter. We computed an F2 score to summarize the algorithm’s performance at differentiating wake and sleep, placing more weight on the algorithm’s sensitivity to capture daytime sleep. Results Twenty-seven individuals (19F/8M; average age 62.33 ± 3.04 years) were recruited from the poststroke inpatient unit of a rehabilitation hospital. The ActiGraph Cole-Kripke algorithm (configured with minimum sleep time=15 mins, bedtime=10 mins, and wake time=10 mins) yielded the highest F2 score (F2=0.59), outperforming Sadeh (F2=0.57) and ActiWatch ARMI (F2=0.52) algorithms under their respective optimized parameters. When exclusively considering data from participants lying in bed, the ActiGraph device consistently achieved superior performance (F2=0.69) with the same optimized Cole-Kripke settings. Conclusion In poststroke patients in an inpatient rehabilitation unit, ActiGraph (Cole-Kripke) was better than ActiWatch in detecting daytime sleep. The results could help inform specific algorithms and parameters that can be readily implemented for daytime sleep monitoring and EDS detection in poststroke patients. Future considerations for enhanced algorithms should include posture detection to maximize efficacy. Support (if any) NIH R01HD097786-01A1.