Novel sleep metrics derived from chest-worn accelerometers are associated with dementia risk: The ARIC Neurocognitive Study
Keran W Chamberlin, Lacey H Etzkorn, Adam P Spira, Kening Jiang, Lin Yee Chen, Ciprian Crainiceanu, Vadim Zipunnikov, Jennifer A SchrackAbstract
Study Objectives
1) To develop an open-source method to identify time in bed (TIB) and related sleep parameters from chest-worn accelerometers; 2) To quantify the associations between derived sleep parameters and dementia risk.
Methods
Sleep parameters were estimated using new methods applied to movement and posture information extracted from the Zio® XT onboard accelerometer worn on the chest by participants in the Atherosclerosis Risk in Communities (ARIC) Study (2016-2017). Methods were (1) calibrated based on published methods applied to data collected simultaneously using hip-worn actigraphy in a subset of the ARIC participants (n=405, age 78.5±4.6 years); and (2) applied to all ARIC participants free of dementia (n=2,316, age 79.0±4.5 years). Cause-specific Cox regressions were used to evaluate associations of these estimated parameters with incident dementia.
Results
Participants contributed an average of 5.7 days to the calibration process. Incorporating postural information from the accelerometer reduced misclassification of sedentary status as TIB, especially during early nighttime hours. During a median 5.2 years (Q1:Q3: 4.5-5.7) of follow-up, 326 participants (14.1%) developed dementia. Both TIB and total sleep time showed U-shaped associations with dementia risk. The lowest tertile of interdaily variability in TIB midpoint was associated with lower dementia hazard. Dementia risk was not significantly associated with TIB midpoint, wake after sleep onset, or sleep efficiency.
Conclusions
Chest-worn accelerometers provide promising estimates of sleep parameters because they combine activity intensity and postural information. Derived TIB, total sleep time, and variability in the timing of TIB across days are associated with dementia risk.