Sleep estimation from wrist movement quantified by different actigraphic modalities

Abstract

Progress in transducer design and empirical characterization of wrist movement has led to diverse wrist activity monitors, each with its unique features and modality of operation. This study compared sleep--wake estimates from nocturnal wrist activity quantified by different motion-quantifying algorithms. Healthy young adults wore an Actillume and a Mini Motionlogger on the same wrist while nocturnal polysomnography data were recorded simultaneously in the laboratory. Activity data were analyzed with ACTION3 using scoring algorithms independently calibrated for each measurement modality. Overall, each modality yielded accurate and reliable sleep estimates relative to polysomnographic estimates (agreement rates: 91.4--96.5%, correlations for sleep duration: 0.79--0.94). Estimates derived from Actillume modalities were comparable to those of Mini Motionloggers, suggesting that the transducers of these two devices performed comparably for monitoring sleep and wakefulness. Wrist movement quantified by the Mini Motionlogger proportional-integrating mode yielded the best accuracy for detection of sleep--wake states.